diff --git a/questions/BallDrop/dataframes/001063c5dfa248dbbba122ff6743af5f.parquet b/questions/BallDrop/dataframes/001063c5dfa248dbbba122ff6743af5f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e3d66c8428483f84ca9c332ee4221ff84eda38e9 --- /dev/null +++ b/questions/BallDrop/dataframes/001063c5dfa248dbbba122ff6743af5f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e002138f60760655121db4940d496df7a4a93d5de9119c669ee356ae35568a8 +size 48831 diff --git a/questions/BallDrop/dataframes/00ae112c60ec4dd1acd1bacbec2560c9.parquet b/questions/BallDrop/dataframes/00ae112c60ec4dd1acd1bacbec2560c9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..abd04b564136f906cb6a3d0f93b17952042bd934 --- /dev/null +++ b/questions/BallDrop/dataframes/00ae112c60ec4dd1acd1bacbec2560c9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d7ef1493084ed4b97fe8180019bb38b3436d0cc4be3895ecc1e84a3adc56826 +size 68834 diff --git a/questions/BallDrop/dataframes/00f96c534483544fbf5cb3d14e38272f.parquet b/questions/BallDrop/dataframes/00f96c534483544fbf5cb3d14e38272f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1aa73bafff9a34f21333a4a762ab81b769ad139e --- /dev/null +++ b/questions/BallDrop/dataframes/00f96c534483544fbf5cb3d14e38272f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:007831fb081137e68933025b794bbaccbb39b1585ed17ffdd10d3ec9f0f921fe +size 42201 diff --git a/questions/BallDrop/dataframes/06c7aeb4badb5ca085c44da26f39cfa5.parquet b/questions/BallDrop/dataframes/06c7aeb4badb5ca085c44da26f39cfa5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..90889c2a730fb503110fa2000dc52bfb8c40f3e4 --- /dev/null +++ b/questions/BallDrop/dataframes/06c7aeb4badb5ca085c44da26f39cfa5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee96eeff4a1a84662399277bde4cbf71089d3037c10332125e8c00c977d1f0ef +size 62995 diff --git a/questions/BallDrop/dataframes/08a23a9f66584605a0ce62ddaf3fcdd8.parquet b/questions/BallDrop/dataframes/08a23a9f66584605a0ce62ddaf3fcdd8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..12d9cc01dce73f94cf99bef2a836e0362da1553d --- /dev/null +++ b/questions/BallDrop/dataframes/08a23a9f66584605a0ce62ddaf3fcdd8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:370315e5bbab3b4361627d087ab5b8555f680dc24e07bff91f44878ac5766df3 +size 49885 diff --git a/questions/BallDrop/dataframes/0d96332e3fae53ff8c19e0445a16f304.parquet b/questions/BallDrop/dataframes/0d96332e3fae53ff8c19e0445a16f304.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ee7235ad7d4aad91b8016837b784b71334a0d917 --- /dev/null +++ b/questions/BallDrop/dataframes/0d96332e3fae53ff8c19e0445a16f304.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb081b6a459bc8868b06c4401c99a5ce91455ad40b84af74b646aae52f9adc9a +size 65241 diff --git a/questions/BallDrop/dataframes/0f159f29c1004403a466c6bea03d7d02.parquet b/questions/BallDrop/dataframes/0f159f29c1004403a466c6bea03d7d02.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c1bfe715d51aad0cc7175aca8fce71936b68f263 --- /dev/null +++ b/questions/BallDrop/dataframes/0f159f29c1004403a466c6bea03d7d02.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:671c0a3153bf60b5662b5cb8d43cf55e7744046d6827c1955432fd5ca3a3d365 +size 65020 diff --git a/questions/BallDrop/dataframes/12468fab1955463b813e11f74847029b.parquet b/questions/BallDrop/dataframes/12468fab1955463b813e11f74847029b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..008ac5ae8f10fae56ec194a1680d3f4141c802ea --- /dev/null +++ b/questions/BallDrop/dataframes/12468fab1955463b813e11f74847029b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93181333e95e08d4c2e09211354d2acd700eabe48e06188c552d692526782089 +size 32143 diff --git a/questions/BallDrop/dataframes/17a033a89614421b8bb520c3bb286b60.parquet b/questions/BallDrop/dataframes/17a033a89614421b8bb520c3bb286b60.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e4594b787450258ad6d88baae62465ca8fb2bc96 --- /dev/null +++ b/questions/BallDrop/dataframes/17a033a89614421b8bb520c3bb286b60.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:487f7a3b8a68b57d7d9b2a3227ab4e40bf514bc9fde8434e12f18a028bf33adf +size 68929 diff --git a/questions/BallDrop/dataframes/18e695472b2d52058cf6deee57c49cf8.parquet b/questions/BallDrop/dataframes/18e695472b2d52058cf6deee57c49cf8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..47f95948489ee5295efa96ea2fb1e69d6089f22c --- /dev/null +++ b/questions/BallDrop/dataframes/18e695472b2d52058cf6deee57c49cf8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:666f60381e3f7ce6cd5d6d5c45f8f032c8ecd1b874b32001f791d3edf5ac21d0 +size 48399 diff --git a/questions/BallDrop/dataframes/1b4abf9677a956e4bb90051b4494f4b3.parquet b/questions/BallDrop/dataframes/1b4abf9677a956e4bb90051b4494f4b3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..66bb6a97a66bbb67f73ce0885d8fb2bb47154663 --- /dev/null +++ b/questions/BallDrop/dataframes/1b4abf9677a956e4bb90051b4494f4b3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61f3ace6688f20bbe67535dcec8586cdf27e92ec1f67f441c2eff43ff87224b6 +size 68880 diff --git a/questions/BallDrop/dataframes/1dcbb406a6365e2f86382f83d5a8a508.parquet b/questions/BallDrop/dataframes/1dcbb406a6365e2f86382f83d5a8a508.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e11644517cefa4444c05e57198ff1fabe798375d --- /dev/null +++ b/questions/BallDrop/dataframes/1dcbb406a6365e2f86382f83d5a8a508.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33c9ae477c2cdb986af8fcd714da596e91015fc5c7375f6ceb561ea3cf507fed +size 35520 diff --git a/questions/BallDrop/dataframes/1f176712882c5ce69f5f87c4473b6694.parquet b/questions/BallDrop/dataframes/1f176712882c5ce69f5f87c4473b6694.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e80d290f78b8b4abe02808b39c027fe6b79a692b --- /dev/null +++ b/questions/BallDrop/dataframes/1f176712882c5ce69f5f87c4473b6694.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b7825802edf18d199b8469db1547f366014d9671e41b8745a5ef92a9738ff98 +size 43484 diff --git a/questions/BallDrop/dataframes/2348ceba66234f5f91acc812e273a3d5.parquet b/questions/BallDrop/dataframes/2348ceba66234f5f91acc812e273a3d5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9b16585ae2ec08501d5bd949277a76d7bed86ca6 --- /dev/null +++ b/questions/BallDrop/dataframes/2348ceba66234f5f91acc812e273a3d5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a29effeec48bb4669071ac07d274ca5349b95425cbd555bed77d28376db8feed +size 41541 diff --git a/questions/BallDrop/dataframes/2711a07092314ea681315ca82e5bc51b.parquet b/questions/BallDrop/dataframes/2711a07092314ea681315ca82e5bc51b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1ad71b8a4dafb31666d36295bd1c1f659c7b7167 --- /dev/null +++ b/questions/BallDrop/dataframes/2711a07092314ea681315ca82e5bc51b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a37671793eb17d4df33c8faa6196841a7df05051a0b33a2e2f3a447c208a4d40 +size 69061 diff --git a/questions/BallDrop/dataframes/2e34f5dce75f46be81d22e0deae7b30f.parquet b/questions/BallDrop/dataframes/2e34f5dce75f46be81d22e0deae7b30f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..81c408f7c62412aed85c1341b3110e3c172e0065 --- /dev/null +++ b/questions/BallDrop/dataframes/2e34f5dce75f46be81d22e0deae7b30f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1e7190b1a522861e4502b682841a8cf7f1dddecd41c39c9c0afc53058c4fb59 +size 68812 diff --git a/questions/BallDrop/dataframes/2e86956a56055b2aa94c0030af2f2f74.parquet b/questions/BallDrop/dataframes/2e86956a56055b2aa94c0030af2f2f74.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1e24c5bd2d60e9af33ee080e3a449dd1a6fb4a2c --- /dev/null +++ b/questions/BallDrop/dataframes/2e86956a56055b2aa94c0030af2f2f74.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0f22aeb6bbee1664a65343b0de580d62937ac7d9af46fd33f695bbe2a574e77 +size 68815 diff --git a/questions/BallDrop/dataframes/303d59b812a942eeac4a05f71df85ff1.parquet b/questions/BallDrop/dataframes/303d59b812a942eeac4a05f71df85ff1.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b36a625da5b5df53a9e4722c52f1bd43d6934c02 --- /dev/null +++ b/questions/BallDrop/dataframes/303d59b812a942eeac4a05f71df85ff1.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7566940db6182dde000ee39232b460188e0a4f2d16eff98ff555b9ab67b39d6e +size 68993 diff --git a/questions/BallDrop/dataframes/33508fdf1d3d5b62a0f61a2ab5200441.parquet b/questions/BallDrop/dataframes/33508fdf1d3d5b62a0f61a2ab5200441.parquet new file mode 100644 index 0000000000000000000000000000000000000000..357a6577968f84997446bffc54075294bb1bef5d --- /dev/null +++ b/questions/BallDrop/dataframes/33508fdf1d3d5b62a0f61a2ab5200441.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf2836b350df2b7f58381f58135bee9fc709419baa66d2276e731a84d127dc9f +size 46142 diff --git a/questions/BallDrop/dataframes/34574575af47569e8859d01ec8078234.parquet b/questions/BallDrop/dataframes/34574575af47569e8859d01ec8078234.parquet new file mode 100644 index 0000000000000000000000000000000000000000..74d9581a7f1a2ba65760b2ea755846e22b3b0419 --- /dev/null +++ b/questions/BallDrop/dataframes/34574575af47569e8859d01ec8078234.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0dc327c73162e3d41b796447bb417a262be179be2437205e72bebd7716ba7341 +size 37288 diff --git a/questions/BallDrop/dataframes/36aee79d125e52d18d0608ce1873c6b4.parquet b/questions/BallDrop/dataframes/36aee79d125e52d18d0608ce1873c6b4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..27cd0debfe2c4eb2b57b617a0251bfa0d2ac8932 --- /dev/null +++ b/questions/BallDrop/dataframes/36aee79d125e52d18d0608ce1873c6b4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02552036ab713a1541a7e74140d68eafcee1eceba43f95a9f0be6848ca8dbfd7 +size 65381 diff --git a/questions/BallDrop/dataframes/39330e4b3b284bfa859687cd7abf9b14.parquet b/questions/BallDrop/dataframes/39330e4b3b284bfa859687cd7abf9b14.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e3b311025c640122b0a70da3ca6f7b2dee9d283a --- /dev/null +++ b/questions/BallDrop/dataframes/39330e4b3b284bfa859687cd7abf9b14.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51aa2d92aa22551b7d90eb86f8a85194760f94e07899c9c0a7272cb5ddb05da2 +size 68833 diff --git a/questions/BallDrop/dataframes/39766784659f583399ef48bbd86c3a85.parquet b/questions/BallDrop/dataframes/39766784659f583399ef48bbd86c3a85.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8ba6d491b7cb60d7a5bc3ec262c26d99a4a01813 --- /dev/null +++ b/questions/BallDrop/dataframes/39766784659f583399ef48bbd86c3a85.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21c9de55bfe7aa22841e4d4328828340fb1472dbadb1f4d102a5bc462062793b +size 53530 diff --git a/questions/BallDrop/dataframes/3a5ad8a0c7a84fd3a5baaec0eee7a126.parquet b/questions/BallDrop/dataframes/3a5ad8a0c7a84fd3a5baaec0eee7a126.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e582ae258f392d761ddfadf494f20b8a0b0c9967 --- /dev/null +++ b/questions/BallDrop/dataframes/3a5ad8a0c7a84fd3a5baaec0eee7a126.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5490c7a9dcc97f4ebe31c9b744ccd0c32f0c181f85491d5775f2f670427c334a +size 33577 diff --git a/questions/BallDrop/dataframes/3df1d923d25b44d6826383b65210f7eb.parquet b/questions/BallDrop/dataframes/3df1d923d25b44d6826383b65210f7eb.parquet new file mode 100644 index 0000000000000000000000000000000000000000..567d9f78de52c63707f0461f65bc99c0f2bf20e2 --- /dev/null +++ b/questions/BallDrop/dataframes/3df1d923d25b44d6826383b65210f7eb.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1261ffacf16aeaa649523b1455dd79e0cc584befc13949f7a696c60b90abfb7 +size 41917 diff --git a/questions/BallDrop/dataframes/3e28f2d025fd4f229ce62a07f1ed72eb.parquet b/questions/BallDrop/dataframes/3e28f2d025fd4f229ce62a07f1ed72eb.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f984d686c332caca3c1792b23f00681bc695e026 --- /dev/null +++ b/questions/BallDrop/dataframes/3e28f2d025fd4f229ce62a07f1ed72eb.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b5c2c9f4234b457ebf28dded2002dfc4aa6f8cb9efbcb9a01d96af47b5e2e84 +size 57175 diff --git a/questions/BallDrop/dataframes/3e706177423c5f289117291cc571bc8d.parquet b/questions/BallDrop/dataframes/3e706177423c5f289117291cc571bc8d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..326e1bace0e6d9c1abf59abf3f32b6284c63ad84 --- /dev/null +++ b/questions/BallDrop/dataframes/3e706177423c5f289117291cc571bc8d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de60c72305f27d75d3ae5695f704c132e0d7bb8773e9c1c70a0e7b47518482a4 +size 64699 diff --git a/questions/BallDrop/dataframes/42e0669504784adea7e79a929f5c5495.parquet b/questions/BallDrop/dataframes/42e0669504784adea7e79a929f5c5495.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2263e1ead2cdfcbb29e8742a50411759485eb9e5 --- /dev/null +++ b/questions/BallDrop/dataframes/42e0669504784adea7e79a929f5c5495.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb795ab2dbe84f177ae59b61002c8b2e3996bf83a423732a086d56f8b94890a3 +size 68861 diff --git a/questions/BallDrop/dataframes/4373eb3bac935c4e8b80d82e28f50ffe.parquet b/questions/BallDrop/dataframes/4373eb3bac935c4e8b80d82e28f50ffe.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e0d52c1d403cb4bd397581e1c010b94118354dd8 --- /dev/null +++ b/questions/BallDrop/dataframes/4373eb3bac935c4e8b80d82e28f50ffe.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b8ac7e6bff82fc35bb348ab44363339c151ccd207f9165d74b7bd5ab478caa6 +size 68849 diff --git a/questions/BallDrop/dataframes/439d72f79dac13bab27ea2190255b070.parquet b/questions/BallDrop/dataframes/439d72f79dac13bab27ea2190255b070.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3f9b114ace201311a872bb0b4ecc050f7ca55ccc --- /dev/null +++ b/questions/BallDrop/dataframes/439d72f79dac13bab27ea2190255b070.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a7463f5d9a944bedacfc5a7fd87a4026ecf87e36e9b6fdd6cf3a0b0a2df3b12 +size 55977 diff --git a/questions/BallDrop/dataframes/4757e7217e7a56d693494ebf2f740d43.parquet b/questions/BallDrop/dataframes/4757e7217e7a56d693494ebf2f740d43.parquet new file mode 100644 index 0000000000000000000000000000000000000000..352e213e1956e450f1993f0217f1bb5d31c9f291 --- /dev/null +++ b/questions/BallDrop/dataframes/4757e7217e7a56d693494ebf2f740d43.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8aac2adc880b6ab02d4d11b83215ca86afd9322ebe8e07bf7f3cbc8d19f760f +size 68839 diff --git a/questions/BallDrop/dataframes/47f84b9f0967540a98380bd737333252.parquet b/questions/BallDrop/dataframes/47f84b9f0967540a98380bd737333252.parquet new file mode 100644 index 0000000000000000000000000000000000000000..de62d39df33aec2ccf42d989a88798fbcb618850 --- /dev/null +++ b/questions/BallDrop/dataframes/47f84b9f0967540a98380bd737333252.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5f5f256624a24c2b8f7958ec19345bf4d268bd20593cecd98579177ca214ec3 +size 69011 diff --git a/questions/BallDrop/dataframes/4c1e9b4e56e32073502d64f3e9b72356.parquet b/questions/BallDrop/dataframes/4c1e9b4e56e32073502d64f3e9b72356.parquet new file mode 100644 index 0000000000000000000000000000000000000000..05ca9984db2cca603303d6f16d37a8c6af120271 --- /dev/null +++ b/questions/BallDrop/dataframes/4c1e9b4e56e32073502d64f3e9b72356.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19699239303e168cb1b97948b1aaff781da538da222b7c099d8cad178f815235 +size 68812 diff --git a/questions/BallDrop/dataframes/4d36b3bd1e524a49b337e2ac9a97206b.parquet b/questions/BallDrop/dataframes/4d36b3bd1e524a49b337e2ac9a97206b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..04e661e5b126fce7e837b1f9cf866e2545303249 --- /dev/null +++ b/questions/BallDrop/dataframes/4d36b3bd1e524a49b337e2ac9a97206b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6309c324ee5765577b2575e93d5dbcb1b0d0b3eb8e5cec12d3ac0d3b7edb2de9 +size 51534 diff --git a/questions/BallDrop/dataframes/50b90a6df0e25ac480a3af80bb1b21bd.parquet b/questions/BallDrop/dataframes/50b90a6df0e25ac480a3af80bb1b21bd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..70af992e8394207ea4ddd5c2cc9029139614d931 --- /dev/null +++ b/questions/BallDrop/dataframes/50b90a6df0e25ac480a3af80bb1b21bd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f40de4d6876b60a9f4754d9b00e2fcf83aecc333cb795d7c67bd60a858a6c1a +size 55033 diff --git a/questions/BallDrop/dataframes/523c9218f52c57068887247f1c7a044a.parquet b/questions/BallDrop/dataframes/523c9218f52c57068887247f1c7a044a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8d66cbe6aaad5c66898ff57cd4876a6d72cb5ab8 --- /dev/null +++ b/questions/BallDrop/dataframes/523c9218f52c57068887247f1c7a044a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a48663da4be4bd54816c222ade04578b53ee6e690d459d07e91f9e7bc4ed5cd0 +size 45193 diff --git a/questions/BallDrop/dataframes/55cd4372b9205e238624bf60157b1250.parquet b/questions/BallDrop/dataframes/55cd4372b9205e238624bf60157b1250.parquet new file mode 100644 index 0000000000000000000000000000000000000000..507ccd03933362742557650147b669cdfad39862 --- /dev/null +++ b/questions/BallDrop/dataframes/55cd4372b9205e238624bf60157b1250.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5cef9fb8966d8877f785787393bfa7d29ac7573178ca2a955bade81617441bf0 +size 35004 diff --git a/questions/BallDrop/dataframes/56329b0d596f5b4daa3db101e5a634e6.parquet b/questions/BallDrop/dataframes/56329b0d596f5b4daa3db101e5a634e6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b5a0c6d3b5914639dc9f643fa411ba350808b7ef --- /dev/null +++ b/questions/BallDrop/dataframes/56329b0d596f5b4daa3db101e5a634e6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:710b1b5bf2d93118d54b5a8943c6f0d2dc8a92ccccedbf0707bd3d202cd316f1 +size 38587 diff --git a/questions/BallDrop/dataframes/568b3e9ed49658d7abaae622daac8cef.parquet b/questions/BallDrop/dataframes/568b3e9ed49658d7abaae622daac8cef.parquet new file mode 100644 index 0000000000000000000000000000000000000000..da7f74efeeb46a4e974153f1c017c990a3012c9f --- /dev/null +++ b/questions/BallDrop/dataframes/568b3e9ed49658d7abaae622daac8cef.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aec21b2dd79e77eb0f7f8387b7950d184b4f2c0f6ae37cc0e50ae282732e8969 +size 51144 diff --git a/questions/BallDrop/dataframes/5f875cc969b55f368b95aee1a44d01e5.parquet b/questions/BallDrop/dataframes/5f875cc969b55f368b95aee1a44d01e5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c66bbf729758324db10b86c1944f4bf13f2b49ae --- /dev/null +++ b/questions/BallDrop/dataframes/5f875cc969b55f368b95aee1a44d01e5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:584dc3e41b089884dfdecde29a0c3b6bb8b8a9136ede8bd8567696d366b840c5 +size 68810 diff --git a/questions/BallDrop/dataframes/61cd70ea73e1497f981510b030571fe5.parquet b/questions/BallDrop/dataframes/61cd70ea73e1497f981510b030571fe5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5fce5afd9b76d129abd2523c95d45a09b512be41 --- /dev/null +++ b/questions/BallDrop/dataframes/61cd70ea73e1497f981510b030571fe5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ad844c91d1344ad46df08e727e14b9bd91ad67b6a40995a880b4dbce40d25ff +size 39614 diff --git a/questions/BallDrop/dataframes/6228a7c919375242b6065fac22112b1e.parquet b/questions/BallDrop/dataframes/6228a7c919375242b6065fac22112b1e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..85d3565d9ea39806ae5a793128abd62139d06309 --- /dev/null +++ b/questions/BallDrop/dataframes/6228a7c919375242b6065fac22112b1e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ee62848089ca379e099efba33e8891ae74fcb82d7ca0ab3511e30fc0f80397a +size 46054 diff --git a/questions/BallDrop/dataframes/631af851dff9494ca7d21f21dac03a3e.parquet b/questions/BallDrop/dataframes/631af851dff9494ca7d21f21dac03a3e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..409ac3726dd9f4d708e783dbec71ffdc2fcb30f6 --- /dev/null +++ b/questions/BallDrop/dataframes/631af851dff9494ca7d21f21dac03a3e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf2bd7e5321cb3aa5c951417f83f6e9b0b35f1c414784088971a0b9c1c3b7b02 +size 57511 diff --git a/questions/BallDrop/dataframes/63b4f438163057f9b67ff2a42d6427c6.parquet b/questions/BallDrop/dataframes/63b4f438163057f9b67ff2a42d6427c6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e7e0d991053f4399f6f80c64d030da675d5147c4 --- /dev/null +++ b/questions/BallDrop/dataframes/63b4f438163057f9b67ff2a42d6427c6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6302c912010e65c9e79e7aebf8067a41abb401750f9b3b28a49612730f738925 +size 41953 diff --git a/questions/BallDrop/dataframes/6514e4c5d16f5fac9810e4b4e0db5d24.parquet b/questions/BallDrop/dataframes/6514e4c5d16f5fac9810e4b4e0db5d24.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a034aa46befddd6a2870010359c98ce76ca6b465 --- /dev/null +++ b/questions/BallDrop/dataframes/6514e4c5d16f5fac9810e4b4e0db5d24.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b52441f6bbb3026507ce290af913270afcb46fe148b5793e942f5abceb5dcdcf +size 42961 diff --git a/questions/BallDrop/dataframes/6c0a4bc64f2f4ff99f76cd41e1ab474d.parquet b/questions/BallDrop/dataframes/6c0a4bc64f2f4ff99f76cd41e1ab474d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9600391df936bfc915740b377323a844dcaf3dfa --- /dev/null +++ b/questions/BallDrop/dataframes/6c0a4bc64f2f4ff99f76cd41e1ab474d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0610acd9507bbadfdfbffbfb304c9bc39c06d449b6b1c039426e6b27d5288d64 +size 69102 diff --git a/questions/BallDrop/dataframes/6c7ac291a99852eea1bd5fb4a619a097.parquet b/questions/BallDrop/dataframes/6c7ac291a99852eea1bd5fb4a619a097.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ed3370c4631a7eb75f7c6aa1bb411c2425aafd25 --- /dev/null +++ b/questions/BallDrop/dataframes/6c7ac291a99852eea1bd5fb4a619a097.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66fe9dabfddf6cf516aafc3afc41e6ddb50d6366b790b8bb7e12912569a89dd4 +size 68928 diff --git a/questions/BallDrop/dataframes/74558dec52105edfb945508fea345732.parquet b/questions/BallDrop/dataframes/74558dec52105edfb945508fea345732.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cd07d93f8c077f784015eca3e8989411e88c77fa --- /dev/null +++ b/questions/BallDrop/dataframes/74558dec52105edfb945508fea345732.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c069ffbe6bc242034409e80a3e36be7e9db56f3fab761b64f4f081b74b8f8fce +size 35543 diff --git a/questions/BallDrop/dataframes/768257b526264466aff05fd9c41ab915.parquet b/questions/BallDrop/dataframes/768257b526264466aff05fd9c41ab915.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c0a263c4cc81dd29ffdf206f9e6f87c2d8cd1a13 --- /dev/null +++ b/questions/BallDrop/dataframes/768257b526264466aff05fd9c41ab915.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f533a8ce0057c4f64e547c672531009db95e789295739242279c5e2d4717ab8 +size 68916 diff --git a/questions/BallDrop/dataframes/77968ae18bca566cbcffba3830a3fece.parquet b/questions/BallDrop/dataframes/77968ae18bca566cbcffba3830a3fece.parquet new file mode 100644 index 0000000000000000000000000000000000000000..901d4d2680f1a2e1e1364e1602cdcaa7794fbd55 --- /dev/null +++ b/questions/BallDrop/dataframes/77968ae18bca566cbcffba3830a3fece.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:118a4e0a29d237bd2111ed35c849ffd12308327841847de3f9152b90177b8e9c +size 48399 diff --git a/questions/BallDrop/dataframes/7d1b3f1a85c049ff9b8bd7c7a18d4041.parquet b/questions/BallDrop/dataframes/7d1b3f1a85c049ff9b8bd7c7a18d4041.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3c82efd018f0fa95bbf1c1a9ad626904807429c0 --- /dev/null +++ b/questions/BallDrop/dataframes/7d1b3f1a85c049ff9b8bd7c7a18d4041.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e11127df101580a3ec7cd98b611b9ae5c4fc46b058f4c8d8ffc1b513c71a6189 +size 42750 diff --git a/questions/BallDrop/dataframes/7df2e9376cf654358d488e51f581b366.parquet b/questions/BallDrop/dataframes/7df2e9376cf654358d488e51f581b366.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e4f6630b19d689b0d3931f502df4d93be3765a55 --- /dev/null +++ b/questions/BallDrop/dataframes/7df2e9376cf654358d488e51f581b366.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1a921a766f65bf8267e7f671470afd55671ced77468a97018f71665f73960030 +size 49913 diff --git a/questions/BallDrop/dataframes/7e0f85ac769c5ab78201f4cdac386d36.parquet b/questions/BallDrop/dataframes/7e0f85ac769c5ab78201f4cdac386d36.parquet new file mode 100644 index 0000000000000000000000000000000000000000..eaddbd2db3ce6e45a247dae4538e8affc6455f9b --- /dev/null +++ b/questions/BallDrop/dataframes/7e0f85ac769c5ab78201f4cdac386d36.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c4bdc95100c47cf5f3b315221a6ce8979109cc61d6c619ff6a359f3516371eeb +size 39078 diff --git a/questions/BallDrop/dataframes/81867907eb7c5370b8f73d2718d8398d.parquet b/questions/BallDrop/dataframes/81867907eb7c5370b8f73d2718d8398d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..33a38f8ebf86e7c14422776f05871b44d2c95d39 --- /dev/null +++ b/questions/BallDrop/dataframes/81867907eb7c5370b8f73d2718d8398d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23386a02f4f4dfebfe7834bc17e119eb3e45cfce3c3b61454725825a99b2f0ca +size 68849 diff --git a/questions/BallDrop/dataframes/81cb67a56e3b53dabdc75ab3749d1a65.parquet b/questions/BallDrop/dataframes/81cb67a56e3b53dabdc75ab3749d1a65.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f8e6fca304eb19041bd981c4b2569b317291f771 --- /dev/null +++ b/questions/BallDrop/dataframes/81cb67a56e3b53dabdc75ab3749d1a65.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:531ec37c311899f2aa9454044c16110b39c06b2620ec2aca62e322b05c49cda9 +size 69045 diff --git a/questions/BallDrop/dataframes/831770be078648aebe505cfc3add6f65.parquet b/questions/BallDrop/dataframes/831770be078648aebe505cfc3add6f65.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bdd6e780ee3df32fac0733d7178686c2e2ed25fd --- /dev/null +++ b/questions/BallDrop/dataframes/831770be078648aebe505cfc3add6f65.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6a2bd8ccea23203e615cf17062c6e39f3ead5bb55bfde1de4d542cf55375902 +size 41102 diff --git a/questions/BallDrop/dataframes/873c2abc53235d5eb408430e7313c0b7.parquet b/questions/BallDrop/dataframes/873c2abc53235d5eb408430e7313c0b7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6306792bee3ece667a8ef7158dff4a3c11d8786a --- /dev/null +++ b/questions/BallDrop/dataframes/873c2abc53235d5eb408430e7313c0b7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a12204968557fdba94304766458152d61e8df5c2499de057745acc4979adbfb3 +size 47460 diff --git a/questions/BallDrop/dataframes/8845ac4101424920bfad9f1eebc21c1c.parquet b/questions/BallDrop/dataframes/8845ac4101424920bfad9f1eebc21c1c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..01be03b0953702ff00f6d69e021d2b479855ba80 --- /dev/null +++ b/questions/BallDrop/dataframes/8845ac4101424920bfad9f1eebc21c1c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9df43a1add53e8698375e8c34441d6bc40c270cef0103d0fcbc0da24cb06281c +size 48915 diff --git a/questions/BallDrop/dataframes/8f251b7ab3515230926f69d9fcd57dd4.parquet b/questions/BallDrop/dataframes/8f251b7ab3515230926f69d9fcd57dd4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b7f5daf20b0206a96650534d18a1ab2c3b06fd5e --- /dev/null +++ b/questions/BallDrop/dataframes/8f251b7ab3515230926f69d9fcd57dd4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31dac8f3b9136288990278261363e0033d78413738f7f8984bf96587ddf1f95f +size 38328 diff --git a/questions/BallDrop/dataframes/8f30e32b7b6b4d04915b710074abb452.parquet b/questions/BallDrop/dataframes/8f30e32b7b6b4d04915b710074abb452.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7f9a0f78c886b263f16afde74e5231de277b31a8 --- /dev/null +++ b/questions/BallDrop/dataframes/8f30e32b7b6b4d04915b710074abb452.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce0298274190fc02a67ba453be03355562f85d04fa115c2b5236e2b539d5bc0a +size 46586 diff --git a/questions/BallDrop/dataframes/919c229adaf44cd68af7629ae8feb882.parquet b/questions/BallDrop/dataframes/919c229adaf44cd68af7629ae8feb882.parquet new file mode 100644 index 0000000000000000000000000000000000000000..670ecbf7df5498f058db5467b906a6a231b63ff0 --- /dev/null +++ b/questions/BallDrop/dataframes/919c229adaf44cd68af7629ae8feb882.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c07554fbdd699f24971aa9238b74408d66c4310672afcd0d66146c30846a04b7 +size 40030 diff --git a/questions/BallDrop/dataframes/91ddb1413c4fbf36279927c5338fc650.parquet b/questions/BallDrop/dataframes/91ddb1413c4fbf36279927c5338fc650.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4d20e187011f1aa2ed92aea1e2f0ad7c3f09023f --- /dev/null +++ b/questions/BallDrop/dataframes/91ddb1413c4fbf36279927c5338fc650.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e0b89792d92b6d9a379da0b879e39e43e31c6e44b73b97a0820bf507779cf86 +size 68850 diff --git a/questions/BallDrop/dataframes/93651e54ac445eba83c9f71cd64421a9.parquet b/questions/BallDrop/dataframes/93651e54ac445eba83c9f71cd64421a9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..61d31619cc04c58b2e9b47a881b8f1a7b9c34170 --- /dev/null +++ b/questions/BallDrop/dataframes/93651e54ac445eba83c9f71cd64421a9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41820bbc5fdbae677f970b3de6f04f9afa856f3c248a6b2477d44db15d34fd14 +size 34203 diff --git a/questions/BallDrop/dataframes/9469a80f76084842933ee512968eadc2.parquet b/questions/BallDrop/dataframes/9469a80f76084842933ee512968eadc2.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0b08b9d57b92d315becf045bdcddcb06f48ba525 --- /dev/null +++ b/questions/BallDrop/dataframes/9469a80f76084842933ee512968eadc2.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35390541cf56d1938e1d35a5ca3ba19228478f0fd0db89e5226e91bca74dba52 +size 44840 diff --git a/questions/BallDrop/dataframes/9583b6b3a5c548c0893c647f3350dd5a.parquet b/questions/BallDrop/dataframes/9583b6b3a5c548c0893c647f3350dd5a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4aa82c309a5647979ea0ab119e84505fbf834243 --- /dev/null +++ b/questions/BallDrop/dataframes/9583b6b3a5c548c0893c647f3350dd5a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0da303e78f7fb2177cc903dbc6631ab82078617ba6116a41b21789267edf0010 +size 68798 diff --git a/questions/BallDrop/dataframes/9c445001c2f75ddb9bd1d47993a664f2.parquet b/questions/BallDrop/dataframes/9c445001c2f75ddb9bd1d47993a664f2.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0a8f5ea1e90314a47918004267f7255629d0a9d5 --- /dev/null +++ b/questions/BallDrop/dataframes/9c445001c2f75ddb9bd1d47993a664f2.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43a6ce7d87bbeef575d8dbf2ece535937e46ec0a4ea4196cead92b156a7100d8 +size 40333 diff --git a/questions/BallDrop/dataframes/9d38efbdeee64affb143c411f6081d62.parquet b/questions/BallDrop/dataframes/9d38efbdeee64affb143c411f6081d62.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ab58e258a468ba6f61e15929a67ddc520ebd60b9 --- /dev/null +++ b/questions/BallDrop/dataframes/9d38efbdeee64affb143c411f6081d62.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49e3c7c62e69625a259d4fe46f98905013b7e3f541c60ae1a76c3b257b9f04e5 +size 50250 diff --git a/questions/BallDrop/dataframes/9daf764b6051540e98b175ec43c551c4.parquet b/questions/BallDrop/dataframes/9daf764b6051540e98b175ec43c551c4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7c6afb1d5558086f0eb3fe014252783cd0f57811 --- /dev/null +++ b/questions/BallDrop/dataframes/9daf764b6051540e98b175ec43c551c4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a5c4ce6a948e9972cb1f5d4fedb39deed7ac5e539bed14ed7663dfc1e6b37ae +size 68893 diff --git a/questions/BallDrop/dataframes/9efdb21c07b5589b9b02b808cfd7d78f.parquet b/questions/BallDrop/dataframes/9efdb21c07b5589b9b02b808cfd7d78f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..611cc6218c54721b7c4c046b150bd7b5874cd0f0 --- /dev/null +++ b/questions/BallDrop/dataframes/9efdb21c07b5589b9b02b808cfd7d78f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d25465877abd8a8274437b02edb930bee2c85b411d54119d9f8f0c1e27a2605 +size 42252 diff --git a/questions/BallDrop/dataframes/a1c583110649c9166d2b8cc4ab82eca6.parquet b/questions/BallDrop/dataframes/a1c583110649c9166d2b8cc4ab82eca6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..86e3c35aa2441a452cc7532a0e14371c916470fe --- /dev/null +++ b/questions/BallDrop/dataframes/a1c583110649c9166d2b8cc4ab82eca6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:592f21bd88586eb78fa6c56aa4f3c8bd07dfe3afcfb5058e1fecf400fa56afba +size 68941 diff --git a/questions/BallDrop/dataframes/a236c7b60f6d489cad9fdc8ef847a425.parquet b/questions/BallDrop/dataframes/a236c7b60f6d489cad9fdc8ef847a425.parquet new file mode 100644 index 0000000000000000000000000000000000000000..51c17dff213f9b5179874f0c7e75233c78da53e3 --- /dev/null +++ b/questions/BallDrop/dataframes/a236c7b60f6d489cad9fdc8ef847a425.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e50e299a7690e7ecc55a705b977c7625ac848a16297a0b02f86439d7df676e11 +size 60386 diff --git a/questions/BallDrop/dataframes/a26263883fa446e49b5b921cfdf3944d.parquet b/questions/BallDrop/dataframes/a26263883fa446e49b5b921cfdf3944d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bf30cc7cabe31ae7c8b6e4e612c79a0c513855d3 --- /dev/null +++ b/questions/BallDrop/dataframes/a26263883fa446e49b5b921cfdf3944d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:172e4ac4f61271c73c81ac7afca192a7e5ecc5bd4d1f4de3236b3abfb4552d53 +size 67337 diff --git a/questions/BallDrop/dataframes/a58f39410c8a58289fe81de42778819a.parquet b/questions/BallDrop/dataframes/a58f39410c8a58289fe81de42778819a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..304dfdcf98109740a25ffd2ade922285e0457d5b --- /dev/null +++ b/questions/BallDrop/dataframes/a58f39410c8a58289fe81de42778819a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4815f3f7685d95b965971225333e1382fdc8812bf3b8a51d95bd74d4604570d2 +size 68798 diff --git a/questions/BallDrop/dataframes/a6e56514770442afb6af22c381df3a0a.parquet b/questions/BallDrop/dataframes/a6e56514770442afb6af22c381df3a0a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4477bf02258fe2bedb3fafcd6315ebd1707029b1 --- /dev/null +++ b/questions/BallDrop/dataframes/a6e56514770442afb6af22c381df3a0a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16cc146e31d9de2dd97f3d4a1f8745f0eb456b629bf5769708052c49de3beee1 +size 40732 diff --git a/questions/BallDrop/dataframes/a7a514d5d4565e85b4f48ad78735021e.parquet b/questions/BallDrop/dataframes/a7a514d5d4565e85b4f48ad78735021e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ccd8ca2d4d965ffad7622a070c949b1ebe55ed21 --- /dev/null +++ b/questions/BallDrop/dataframes/a7a514d5d4565e85b4f48ad78735021e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f576fc8789d42712a53c9f2415f8ee2096b546e363ca621b86a32cf97d1ada0 +size 68849 diff --git a/questions/BallDrop/dataframes/ab2618c21bea42a5b9d58b93173c082c.parquet b/questions/BallDrop/dataframes/ab2618c21bea42a5b9d58b93173c082c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f2611e1dbf70255a0296203a06dda5faf090d2b3 --- /dev/null +++ b/questions/BallDrop/dataframes/ab2618c21bea42a5b9d58b93173c082c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:528bebbedd3331b6f24e6493a513a654a8ec1eb16ac95cda0aba54ade5dda223 +size 38782 diff --git a/questions/BallDrop/dataframes/adfc9499d2b940ef845993221934f87a.parquet b/questions/BallDrop/dataframes/adfc9499d2b940ef845993221934f87a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5d934c2c42db096508899680da6cbd3ea9b4e983 --- /dev/null +++ b/questions/BallDrop/dataframes/adfc9499d2b940ef845993221934f87a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ab9681960afa20b7fa917ed33f881331698fefa61162347179dbab783d98676 +size 43357 diff --git a/questions/BallDrop/dataframes/b5641f9340fa4cd08749657345a1bedf.parquet b/questions/BallDrop/dataframes/b5641f9340fa4cd08749657345a1bedf.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7ac58c612e1d51b726aece9c44b7d3e998446bf2 --- /dev/null +++ b/questions/BallDrop/dataframes/b5641f9340fa4cd08749657345a1bedf.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8bde909908443937944ae5c6ca2410d8c94d46d7ba82e9f8b69f9ce89e80304 +size 55482 diff --git a/questions/BallDrop/dataframes/b587da7834684a6a9513f7e7b90e8a9e.parquet b/questions/BallDrop/dataframes/b587da7834684a6a9513f7e7b90e8a9e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1b85020f5a7a86d6b2e3408dc5b303d67b72d0df --- /dev/null +++ b/questions/BallDrop/dataframes/b587da7834684a6a9513f7e7b90e8a9e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ff3270ebe9ae55b810885c9b522b2f3cb0a82da9c7d963c1bfd11031aa69706 +size 45624 diff --git a/questions/BallDrop/dataframes/b7c15bc7416f4169b9f2a7166c4865e3.parquet b/questions/BallDrop/dataframes/b7c15bc7416f4169b9f2a7166c4865e3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f059edde2a2e3e69496d76a073e8dfa92600deea --- /dev/null +++ b/questions/BallDrop/dataframes/b7c15bc7416f4169b9f2a7166c4865e3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:259356be25b90711bd07029a6e28fa6eddfd967875f8e67c5162fb83eb19ce7f +size 69029 diff --git a/questions/BallDrop/dataframes/ba2e3cbfe88c52748c6d2c81c7a04600.parquet b/questions/BallDrop/dataframes/ba2e3cbfe88c52748c6d2c81c7a04600.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4b77397772866ee9f50b281fffe955c706bb187f --- /dev/null +++ b/questions/BallDrop/dataframes/ba2e3cbfe88c52748c6d2c81c7a04600.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b61a07aa4ab0fb577d708aeafd07e61f64536fe44bfaeadd79b1bd08bd13fce +size 48399 diff --git a/questions/BallDrop/dataframes/bb0abe5b86f652bf819eb1c65190c4c7.parquet b/questions/BallDrop/dataframes/bb0abe5b86f652bf819eb1c65190c4c7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..98979d2e3392cd0b081fb102d26db049b6b1884d --- /dev/null +++ b/questions/BallDrop/dataframes/bb0abe5b86f652bf819eb1c65190c4c7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6f30bdf2d05d664ff58833d9dd782e4fe4d0f516bc085ca26f9054244199281 +size 33792 diff --git a/questions/BallDrop/dataframes/bc7a8ee50c0b4cd3bf478e3b3f4230dd.parquet b/questions/BallDrop/dataframes/bc7a8ee50c0b4cd3bf478e3b3f4230dd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..76a7f990d15e3db7d72c26150ec4392bfa84763e --- /dev/null +++ b/questions/BallDrop/dataframes/bc7a8ee50c0b4cd3bf478e3b3f4230dd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c2e159c020d2a8cca669a5ae6ef8e7bcf7a976cb4e8fe71ac8ec003c767f19e +size 69046 diff --git a/questions/BallDrop/dataframes/c3eb6b6c9abe403faec7dd1a28b4378e.parquet b/questions/BallDrop/dataframes/c3eb6b6c9abe403faec7dd1a28b4378e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9e1669584328535b7b738a8fa0edd50973a5c726 --- /dev/null +++ b/questions/BallDrop/dataframes/c3eb6b6c9abe403faec7dd1a28b4378e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e2cedb184f3d3911891a6f306489c326a7818c2b5053ac9720f5f7093cecabe6 +size 68839 diff --git a/questions/BallDrop/dataframes/c526889cd0b44bdb9efc5223a2d024a6.parquet b/questions/BallDrop/dataframes/c526889cd0b44bdb9efc5223a2d024a6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..68c536e5fb72094a1fc85577d75add1e904c3eff --- /dev/null +++ b/questions/BallDrop/dataframes/c526889cd0b44bdb9efc5223a2d024a6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfb0e25f0e0c0b703592fb506da3b1056a29f4ea8395ec2dc4e1d66d4d5e4ebc +size 35545 diff --git a/questions/BallDrop/dataframes/c5333739516d5fbfb8aa22aaf11a7cc9.parquet b/questions/BallDrop/dataframes/c5333739516d5fbfb8aa22aaf11a7cc9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e412084890c6a802e9d2e5b552253558e62c7013 --- /dev/null +++ b/questions/BallDrop/dataframes/c5333739516d5fbfb8aa22aaf11a7cc9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d334bf5a233d592557b587d873a26ba674cc323e26b60c2f56904b439acb2335 +size 38980 diff --git a/questions/BallDrop/dataframes/c8baff863e805d089948e883182b8d93.parquet b/questions/BallDrop/dataframes/c8baff863e805d089948e883182b8d93.parquet new file mode 100644 index 0000000000000000000000000000000000000000..43046fa2baf4ba740a1eb4b139ab7f1341500d89 --- /dev/null +++ b/questions/BallDrop/dataframes/c8baff863e805d089948e883182b8d93.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e7985089ae658f55182d0d99afb49f893d15d968b10146750b047ad85223d72 +size 31856 diff --git a/questions/BallDrop/dataframes/ca1727e8ff895ff497d01835818489e8.parquet b/questions/BallDrop/dataframes/ca1727e8ff895ff497d01835818489e8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..39763921fe7de7f22072dfc2980ce7d650009e42 --- /dev/null +++ b/questions/BallDrop/dataframes/ca1727e8ff895ff497d01835818489e8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:037b530747c4aa4d85f69433c8a8eec5a519eac9a435f9771b05c4d14d20c083 +size 47865 diff --git a/questions/BallDrop/dataframes/ca218ab4f28747f6b38fc0149e36eb9b.parquet b/questions/BallDrop/dataframes/ca218ab4f28747f6b38fc0149e36eb9b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ede27b340c05bc1a5b81ef687afb39d2e89b2d84 --- /dev/null +++ b/questions/BallDrop/dataframes/ca218ab4f28747f6b38fc0149e36eb9b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc40f999bed98cebc27ef78d4da4f8f91b3251bcbf6822944e17a672d3e7bd0f +size 36834 diff --git a/questions/BallDrop/dataframes/ccf1034ad54459debd65aff402728604.parquet b/questions/BallDrop/dataframes/ccf1034ad54459debd65aff402728604.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8677cfde239480824f18412e2610072e3e725888 --- /dev/null +++ b/questions/BallDrop/dataframes/ccf1034ad54459debd65aff402728604.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:290c054bb8015cb08352502c91dbdd3dcd8ac57dd9d3e877c65cd1fef417b2a8 +size 66628 diff --git a/questions/BallDrop/dataframes/cd023a4ad4de4316bba0e94b5bf35d8c.parquet b/questions/BallDrop/dataframes/cd023a4ad4de4316bba0e94b5bf35d8c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..12c06e97dcd42596b285c4c21f68dcd4e649d722 --- /dev/null +++ b/questions/BallDrop/dataframes/cd023a4ad4de4316bba0e94b5bf35d8c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c354c23121f9a37c08d583167cc2e338bb6de6c2ba9f3f9185cacb155bcdfa54 +size 68833 diff --git a/questions/BallDrop/dataframes/cf5492d1ceb754e6af4c83df1d1e646f.parquet b/questions/BallDrop/dataframes/cf5492d1ceb754e6af4c83df1d1e646f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..374b85bd4cf7e53ad2eb545a8539122d20e74728 --- /dev/null +++ b/questions/BallDrop/dataframes/cf5492d1ceb754e6af4c83df1d1e646f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66c91b2c6ef44005602236f8536ec94313d235ce11059d6eb3ce07af99758586 +size 65073 diff --git a/questions/BallDrop/dataframes/cf90c926e3d14e669525f9577984bec0.parquet b/questions/BallDrop/dataframes/cf90c926e3d14e669525f9577984bec0.parquet new file mode 100644 index 0000000000000000000000000000000000000000..63d27356563f2a385821ac171c0d12aaab049d78 --- /dev/null +++ b/questions/BallDrop/dataframes/cf90c926e3d14e669525f9577984bec0.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f704b71da491561c8fda29cd5915f8230e4e740fc358f9c77e79e9dbe005900 +size 68839 diff --git a/questions/BallDrop/dataframes/cfc1dde0ab3fe2803be3d63dce48d931.parquet b/questions/BallDrop/dataframes/cfc1dde0ab3fe2803be3d63dce48d931.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7070540a5b01f94954187ff51f622dd4f4bb92fa --- /dev/null +++ b/questions/BallDrop/dataframes/cfc1dde0ab3fe2803be3d63dce48d931.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8caf007f4594d9fcb8f61cc68566a161e438bec09accc886659b6e2d0114db0d +size 68942 diff --git a/questions/BallDrop/dataframes/d5e208a1a02e48d4ba8d47f9cf10a514.parquet b/questions/BallDrop/dataframes/d5e208a1a02e48d4ba8d47f9cf10a514.parquet new file mode 100644 index 0000000000000000000000000000000000000000..169a4592455ab1ee6bbee702b83596acc887120d --- /dev/null +++ b/questions/BallDrop/dataframes/d5e208a1a02e48d4ba8d47f9cf10a514.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:449c092d7e8ed0a1bfa9c760dd4a1dd8d30116fcc5a1971f02178e3f7c7cc326 +size 30701 diff --git a/questions/BallDrop/dataframes/d5e619d9d7fb52898940b8140f653c12.parquet b/questions/BallDrop/dataframes/d5e619d9d7fb52898940b8140f653c12.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6d9629a478fe0d94c222d4cfeb1a293fb7d2bef1 --- /dev/null +++ b/questions/BallDrop/dataframes/d5e619d9d7fb52898940b8140f653c12.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04bfbaa97751ae23ffebbfb421163331ff19c85a4f5e1b895a60de2c7d2c1d4b +size 36411 diff --git a/questions/BallDrop/dataframes/d72e4ead5b465703b756c36d29213e3f.parquet b/questions/BallDrop/dataframes/d72e4ead5b465703b756c36d29213e3f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8ad02b141876697f06e42ea4af1ed51f6bcdd3e8 --- /dev/null +++ b/questions/BallDrop/dataframes/d72e4ead5b465703b756c36d29213e3f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55334fec30b705a1cc2fc1f148209770a43d8e87a0f2cb82cd50535eaa93b602 +size 68930 diff --git a/questions/BallDrop/dataframes/dcbbcb9aaba84f95b6e6c410a573039a.parquet b/questions/BallDrop/dataframes/dcbbcb9aaba84f95b6e6c410a573039a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f07e87e61f2974b597009277b1020ec520f9be31 --- /dev/null +++ b/questions/BallDrop/dataframes/dcbbcb9aaba84f95b6e6c410a573039a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93ddbfb485294465fcf8df3b2f1a174e205e3a9418d9e4872d1ccbab7d401ea4 +size 31870 diff --git a/questions/BallDrop/dataframes/e195837b1041539c86a3733e4391198b.parquet b/questions/BallDrop/dataframes/e195837b1041539c86a3733e4391198b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..22d12449545f5629b5df644230a6a48c015a24c6 --- /dev/null +++ b/questions/BallDrop/dataframes/e195837b1041539c86a3733e4391198b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5104ba8245aa706d24c878e33527db74948f673340594c55b2a455ed556642f3 +size 43765 diff --git a/questions/BallDrop/dataframes/e415e80bfb684ba0a7eec7c7a02ce180.parquet b/questions/BallDrop/dataframes/e415e80bfb684ba0a7eec7c7a02ce180.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c40febd775cc3cef66747b2e983fa925759745c5 --- /dev/null +++ b/questions/BallDrop/dataframes/e415e80bfb684ba0a7eec7c7a02ce180.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aea692bc5830724f4beeb6406a149fd8653b19fff6b27c7f52063448e658293f +size 46346 diff --git a/questions/BallDrop/dataframes/e57e705e527d514db713ff28280d11fd.parquet b/questions/BallDrop/dataframes/e57e705e527d514db713ff28280d11fd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cf73177dedd3d11bf0d055b187c2f1f923aa8951 --- /dev/null +++ b/questions/BallDrop/dataframes/e57e705e527d514db713ff28280d11fd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:286081aa77ef86fe9b9621939957258dc6ab94293a5fb251b721aa1022d312b5 +size 68854 diff --git a/questions/BallDrop/dataframes/e799ab902f045c1d9c29b64ced889454.parquet b/questions/BallDrop/dataframes/e799ab902f045c1d9c29b64ced889454.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b466e00334a02c448c2204a763c4eb35e711bb0e --- /dev/null +++ b/questions/BallDrop/dataframes/e799ab902f045c1d9c29b64ced889454.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c47b99bbacc25eb6e58cb1d3d6b4dd6584974f66a572914b3c75c24577f84957 +size 68839 diff --git a/questions/BallDrop/dataframes/efeae73360874aefb8d1d01031db168a.parquet b/questions/BallDrop/dataframes/efeae73360874aefb8d1d01031db168a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b4534e1ec5ce027ac9c871c6bab67a939e242650 --- /dev/null +++ b/questions/BallDrop/dataframes/efeae73360874aefb8d1d01031db168a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ace008e88f7ad808543c940faa2b710ce9041e9def65378d69dce628a35c82a7 +size 51819 diff --git a/questions/BallDrop/dataframes/f4c8f20b5816534db655784eb8d032f6.parquet b/questions/BallDrop/dataframes/f4c8f20b5816534db655784eb8d032f6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0f2e7761676b52137b677d3301550ecf26b31680 --- /dev/null +++ b/questions/BallDrop/dataframes/f4c8f20b5816534db655784eb8d032f6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19e2f89188e13a0ada1bfab02e58722f9ba30038ac3af8c3d90f1d77fc3dff59 +size 33700 diff --git a/questions/BallDrop/dataframes/f73e9eeb29be5e07b8c13705ff2ab2a6.parquet b/questions/BallDrop/dataframes/f73e9eeb29be5e07b8c13705ff2ab2a6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..db410cb3a7c3465a33f70d7e1cef76992d378022 --- /dev/null +++ b/questions/BallDrop/dataframes/f73e9eeb29be5e07b8c13705ff2ab2a6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f08cd4673f3e85e03665ef9e8dbe5e2e2d1fc6ba44fbeac444bfd523e05ca710 +size 36167 diff --git a/questions/BallDrop/dataframes/f80e701fc3124cb1bf1cf9a5dfb89792.parquet b/questions/BallDrop/dataframes/f80e701fc3124cb1bf1cf9a5dfb89792.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8ae9a35582144b35219d544202bfed428934f63b --- /dev/null +++ b/questions/BallDrop/dataframes/f80e701fc3124cb1bf1cf9a5dfb89792.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2e442a4600edb9f3376253fefedc47eb8379187981458149abb7f5ef2562a93 +size 68832 diff --git a/questions/BallDrop/dataframes/fbe454d4b3625caaaa07aaed18d16268.parquet b/questions/BallDrop/dataframes/fbe454d4b3625caaaa07aaed18d16268.parquet new file mode 100644 index 0000000000000000000000000000000000000000..947b3e9f21ca37c0d4c73a1b405c92245737e70e --- /dev/null +++ b/questions/BallDrop/dataframes/fbe454d4b3625caaaa07aaed18d16268.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:931b7aa8fcab5af2435a3c8d4badaa2cbee334e9f4df053277fa4e077cbe64ea +size 37780 diff --git a/questions/BallDrop/dataframes/fbee8e4e80435cb08b5438b33fb62e36.parquet b/questions/BallDrop/dataframes/fbee8e4e80435cb08b5438b33fb62e36.parquet new file mode 100644 index 0000000000000000000000000000000000000000..722ccffa034bb7051daf6d8a08842b0bfe88d30d --- /dev/null +++ b/questions/BallDrop/dataframes/fbee8e4e80435cb08b5438b33fb62e36.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af7b786cefb3496a80da3b3d3a4140e7ef0da70d788f66a08f94d3ce0b6d4aba +size 66093 diff --git a/questions/BallDrop/dataframes/fca8a26ed97e46f5a04e19c22db27fa5.parquet b/questions/BallDrop/dataframes/fca8a26ed97e46f5a04e19c22db27fa5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4276a32c5255e44e4e73f0e05fa55436c3e66692 --- /dev/null +++ b/questions/BallDrop/dataframes/fca8a26ed97e46f5a04e19c22db27fa5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88d4eab8f1f1d73243e684825fe1bc092a0aa406195d722f8c2d101c7511484e +size 52997 diff --git a/questions/BallDrop/model_record.json b/questions/BallDrop/model_record.json new file mode 100644 index 0000000000000000000000000000000000000000..b453386ecdf3507c9418894bba4c07db976e3b98 --- /dev/null +++ b/questions/BallDrop/model_record.json @@ -0,0 +1,3836 @@ +{ + "4d36b3bd1e524a49b337e2ac9a97206b": { + "parameters_hash": "3071d8a2f38e3bf23b51b9d74fc88ba7", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:21:07.390978", + "end_time_simulation": 9.9975004196167 + }, + "159a715b60519622584f696a929b33d5": { + "parameters_hash": "9d3a353257968b14de0ed07a1c72f4a4", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:21:11.087541", + "end_time_simulation": 9.9975004196167 + }, + "4602bb3dbdef4f5aaf18a3676b46232e": { + "parameters_hash": "2ac7227f2f9ca99ed183e3b1db6a54e4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:13.654332", + "end_time_simulation": 9.9975004196167 + }, + "0d97004813f4491aaedf3afe6893373f": { + "parameters_hash": "a479c6b8f98690bd5e410f3e0aa96b84", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:21:16.070976", + "end_time_simulation": 9.9975004196167 + }, + "61727d0d9fcb429aa1241a1b7925980c": { + "parameters_hash": "7a459aa8a6b40bb30c80e90c47d0ea8e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:18.215447", + "end_time_simulation": 9.9975004196167 + }, + "00ae112c60ec4dd1acd1bacbec2560c9": { + "parameters_hash": "f8a979463f10122d61d65d4294e0c4ed", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:21:20.487204", + "end_time_simulation": 9.9975004196167 + }, + "25726a121f564c52a2837df49d90be3b": { + "parameters_hash": "c58042543e4d1ccc5b4501587a0bc38e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:22.535164", + "end_time_simulation": 9.9975004196167 + }, + "61cd70ea73e1497f981510b030571fe5": { + "parameters_hash": "4d9199aeba0c49ff1830c314aca3993d", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:21:24.787106", + "end_time_simulation": 9.9975004196167 + }, + "1369f277876c4887a9cb7da986d19ddb": { + "parameters_hash": "f26c4506390f6e964c7016d23f88b233", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:26.824525", + "end_time_simulation": 9.9975004196167 + }, + "8f87f89865e44542aaec614f2d21cc84": { + "parameters_hash": "d9e1213509322f9669d1b015b33f5faf", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:21:29.109453", + "end_time_simulation": 9.9975004196167 + }, + "3f3052afbb6d4560855aa43c1b887977": { + "parameters_hash": "c6c6507c4fbd8bdc0a4b31c87252d2e5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:31.171390", + "end_time_simulation": 9.9975004196167 + }, + "b6f90295752f44209d533b6b652360d6": { + "parameters_hash": "d4a4c623c640f10e587b56b98a92d330", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:21:37.459872", + "end_time_simulation": 9.9975004196167 + }, + "439d72f79dac13bab27ea2190255b070": { + "parameters_hash": "3466725c4ee603ab1f14711b20fa9779", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T12:39:15.590619", + "end_time_simulation": 9.9975004196167 + }, + "893bfb84bc29483e97b9ee5061ff0907": { + "parameters_hash": "82e8b0632575b727e6a059d62d779006", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T12:39:18.989427", + "end_time_simulation": 9.9975004196167 + }, + "adfc9499d2b940ef845993221934f87a": { + "parameters_hash": "0a2345ea9190e12b2e5e36189f82e344", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:23:54.069381", + "end_time_simulation": 9.9975004196167 + }, + "6812a4e7e08d5172a53c078eaeebe4c0": { + "parameters_hash": "c05b1fc673cca37047579f8140a4106c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:23:58.244225", + "end_time_simulation": 9.9975004196167 + }, + "ca218ab4f28747f6b38fc0149e36eb9b": { + "parameters_hash": "c150a28c0f7311280884ac76e0927c20", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:21:48.202084", + "end_time_simulation": 9.9975004196167 + }, + "d49134500e0d43d380a13bcd5fa39083": { + "parameters_hash": "50efbe8eb5e745ad570f515c7936b568", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:50.204385", + "end_time_simulation": 9.9975004196167 + }, + "c526889cd0b44bdb9efc5223a2d024a6": { + "parameters_hash": "288d2fd25cfba23511dc57230f2d8302", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:21:52.403759", + "end_time_simulation": 9.9975004196167 + }, + "8502c196042d407d94bc0cd036433448": { + "parameters_hash": "6cb48f5e30d2b9c68e815b3ec8493fca", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:54.408813", + "end_time_simulation": 9.9975004196167 + }, + "e415e80bfb684ba0a7eec7c7a02ce180": { + "parameters_hash": "2a35f33aaeb12783952a0772c11358f7", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:21:56.611810", + "end_time_simulation": 9.9975004196167 + }, + "bb756103727c5070bd82c7e204144f2e": { + "parameters_hash": "a5948e408f9396bfadf13a5a318c2de7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:21:58.656614", + "end_time_simulation": 9.9975004196167 + }, + "2e34f5dce75f46be81d22e0deae7b30f": { + "parameters_hash": "43c806fe8735b2e4aad52cb1ebc76c9b", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:22:00.637508", + "end_time_simulation": 9.9975004196167 + }, + "4c1e9b4e56e32073502d64f3e9b72356": { + "parameters_hash": "53a21a7cc3cebd2f07f8fc35d9f09b16", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T12:39:22.848831", + "end_time_simulation": 9.9975004196167 + }, + "85da477741234e9bb815862c0f52cb0f": { + "parameters_hash": "433b5fe524c1467e60ed3ffe629fa80a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T12:39:26.220369", + "end_time_simulation": 9.9975004196167 + }, + "08a23a9f66584605a0ce62ddaf3fcdd8": { + "parameters_hash": "1d2aa8546042857ed6af7d0cb3c8f281", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:22:07.083965", + "end_time_simulation": 9.9975004196167 + }, + "c68424a8ff2541979d63f2e1a469f33b": { + "parameters_hash": "e751b13d0baf588b694e388e49b8cb23", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:09.119649", + "end_time_simulation": 9.9975004196167 + }, + "85e7a724c1f84b22a56b90e6c3a45096": { + "parameters_hash": "506a6bb705e857c7dd4865ac464ae33a", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:22:11.361225", + "end_time_simulation": 9.9975004196167 + }, + "245824d4141d494faa77b2aaa232115b": { + "parameters_hash": "2604d8014b6f513e1a2a7527d20ee45a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:13.361362", + "end_time_simulation": 9.9975004196167 + }, + "39330e4b3b284bfa859687cd7abf9b14": { + "parameters_hash": "2d43c540013e4b679cbd6eead7f84c80", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:22:15.537933", + "end_time_simulation": 9.9975004196167 + }, + "9a966a7323864bb5884661f9f1987abd": { + "parameters_hash": "a5021389b6bb0020f2c41c60b2515c3b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:17.581648", + "end_time_simulation": 9.9975004196167 + }, + "0f159f29c1004403a466c6bea03d7d02": { + "parameters_hash": "81edab40626ea7642f819abeeedf80c7", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:22:19.564035", + "end_time_simulation": 9.9975004196167 + }, + "91ddb1413c4fbf36279927c5338fc650": { + "parameters_hash": "4274cf48806e11de56f279d9e112f5e3", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:22:21.727396", + "end_time_simulation": 9.9975004196167 + }, + "62bfd011542c4273963548cf624bfa6a": { + "parameters_hash": "7cee3e4bdbbfa056b81624dde44de015", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:23.804681", + "end_time_simulation": 9.9975004196167 + }, + "8845ac4101424920bfad9f1eebc21c1c": { + "parameters_hash": "5cabff5e2d8d11c28d26f779785f0f08", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:22:25.995894", + "end_time_simulation": 9.9975004196167 + }, + "b197f0290372481c961b8ff4cd45f332": { + "parameters_hash": "692707ec403e7f174de0630a94131d24", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:28.010805", + "end_time_simulation": 9.9975004196167 + }, + "9d38efbdeee64affb143c411f6081d62": { + "parameters_hash": "2382887fb18a1662b1f5539350b236af", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:22:30.208979", + "end_time_simulation": 9.9975004196167 + }, + "712c22d9318c427c9f34461035e0a29a": { + "parameters_hash": "e1023e985ff6704b238c3485b64db047", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:32.222860", + "end_time_simulation": 9.9975004196167 + }, + "bc7a8ee50c0b4cd3bf478e3b3f4230dd": { + "parameters_hash": "6891d39b1dff50b72cbb305d4285b096", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:22:34.465089", + "end_time_simulation": 9.9975004196167 + }, + "7475f224222f4c77b0d57222489a0753": { + "parameters_hash": "db08a2b65b1f23db5ad3b086c3e25315", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:36.540383", + "end_time_simulation": 9.9975004196167 + }, + "cf90c926e3d14e669525f9577984bec0": { + "parameters_hash": "a8fc50fcb1a8d2d50b71e061d012c60a", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:22:38.569903", + "end_time_simulation": 9.9975004196167 + }, + "914ab10a0f9947ecac545a7b9b5a9acb": { + "parameters_hash": "df7a983db80828f72671ab3ae3ad42ff", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:22:40.746527", + "end_time_simulation": 9.9975004196167 + }, + "293035ea677e59b29fb79a751ce8e247": { + "parameters_hash": "a2a2f60caf2afdf3e9a59476c554e097", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:42.780755", + "end_time_simulation": 9.9975004196167 + }, + "a26263883fa446e49b5b921cfdf3944d": { + "parameters_hash": "bf3087c0d4645277b415dc4e9dcdb788", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:22:45.012927", + "end_time_simulation": 9.9975004196167 + }, + "842254cc7f275556b3aecd22242db59f": { + "parameters_hash": "ad8ddedbfb3d6fbc07dd9eac0ebd3d19", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:47.051640", + "end_time_simulation": 9.9975004196167 + }, + "8f30e32b7b6b4d04915b710074abb452": { + "parameters_hash": "565f59a9c3a6aa158a5f44abda8371a6", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:22:49.222359", + "end_time_simulation": 9.9975004196167 + }, + "ec112bb5471940b7bd5df0841de15336": { + "parameters_hash": "96506d1d7cc363e1a85c06b75d53c799", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:51.232597", + "end_time_simulation": 9.9975004196167 + }, + "89ac4e4644d24612a511152d378832f1": { + "parameters_hash": "f9adc5ae5a471db8285f1db42eec1937", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:22:53.444060", + "end_time_simulation": 9.9975004196167 + }, + "78b95a0200805d89990cfcdb1b8ccb57": { + "parameters_hash": "389fcc8aca69c7dffb1c5ada58149536", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:22:55.484614", + "end_time_simulation": 9.9975004196167 + }, + "ea78e1117b7c4e0c8257a07ca20ce3a7": { + "parameters_hash": "35a03bd2c86a1a64b769a1d84976c289", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:22:57.531271", + "end_time_simulation": 9.9975004196167 + }, + "cfc1dde0ab3fe2803be3d63dce48d931": { + "parameters_hash": "819df5dc57387bd138a5a6b2f731d574", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:22:59.781338", + "end_time_simulation": 9.9975004196167 + }, + "0e1a49d59cab45b388d69aa0723cedde": { + "parameters_hash": "6b7ac02ca33e0ef943b7845f0b3f5c83", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:01.842860", + "end_time_simulation": 9.9975004196167 + }, + "6c0a4bc64f2f4ff99f76cd41e1ab474d": { + "parameters_hash": "2fe840cf37fbab215afa443314133d86", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:02.267274", + "end_time_simulation": 9.9975004196167 + }, + "72f1263b97f05ef19b5b659d7af651df": { + "parameters_hash": "b210680d7564c4d68b3a0c413d1b622b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:05.705538", + "end_time_simulation": 9.9975004196167 + }, + "631af851dff9494ca7d21f21dac03a3e": { + "parameters_hash": "a738a4f972ff9380e15061c67b7b68df", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:23:08.432107", + "end_time_simulation": 9.9975004196167 + }, + "250d2d57c45c4acdafa81effe6dc2ed3": { + "parameters_hash": "177f203c19306c52d573c108d09162b1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:10.449601", + "end_time_simulation": 9.9975004196167 + }, + "f4dd9a2bec0b4738b60e5df230c36033": { + "parameters_hash": "1e0468810c6933ebdb199fca33dacb9c", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:23:12.678473", + "end_time_simulation": 9.9975004196167 + }, + "ef0b2f9ad39a5e36be60ad2080781e67": { + "parameters_hash": "6e37c79d91d023b349d1be7c3fde428a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:14.663158", + "end_time_simulation": 9.9975004196167 + }, + "b7c15bc7416f4169b9f2a7166c4865e3": { + "parameters_hash": "04f9d1b0727010f396fb1684a5c41598", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:23:16.904164", + "end_time_simulation": 9.9975004196167 + }, + "8ea9c8749d735a8385f6a6ae2f277973": { + "parameters_hash": "d78d997c3cbfb515ea55be7a1a582626", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:18.976601", + "end_time_simulation": 9.9975004196167 + }, + "768257b526264466aff05fd9c41ab915": { + "parameters_hash": "63c775e5f88ff8d94dfea59a5b6eb0ea", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:23:21.176071", + "end_time_simulation": 9.9975004196167 + }, + "060804538b7c42faa451b30b3de7010a": { + "parameters_hash": "3abba47258caa06d30f7c0b5edbcd2c5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:23.238477", + "end_time_simulation": 9.9975004196167 + }, + "03f1084bfb7149eaa6455d85da755ae7": { + "parameters_hash": "3eab57ef8c2354abef8985f8bd0d611b", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "failed", + "timestamp": "2026-05-01T02:46:27.202323", + "error": "\"Unknown BallDrop feature 'restitution_estimate'.\"\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1366, in _compute_feature_from_context\n compute = _FEATURE_COMPUTERS[str(feature_name)]\nKeyError: 'restitution_estimate'\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1282, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 879, in simulate_recipe\n feature_dict = compute_problem_specific_features(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 539, in compute_problem_specific_features\n fn(all_signal_simulation, feature_names=requested_names)\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1448, in compute_features\n return {\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1449, in \n feature_name: _compute_feature_from_context(context, feature_name)\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1368, in _compute_feature_from_context\n raise KeyError(f\"Unknown BallDrop feature '{feature_name}'.\") from exc\nKeyError: \"Unknown BallDrop feature 'restitution_estimate'.\"\n" + }, + "b13c1ce73cecae86b9e01a0556a66164": { + "parameters_hash": "482f35b042f49ff660b4d714e5d0afaf", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "failed", + "timestamp": "2026-05-01T02:46:30.678758", + "error": "\"Unknown BallDrop feature 'restitution_estimate'.\"\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1366, in _compute_feature_from_context\n compute = _FEATURE_COMPUTERS[str(feature_name)]\nKeyError: 'restitution_estimate'\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1282, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 879, in simulate_recipe\n feature_dict = compute_problem_specific_features(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 539, in compute_problem_specific_features\n fn(all_signal_simulation, feature_names=requested_names)\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1448, in compute_features\n return {\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1449, in \n feature_name: _compute_feature_from_context(context, feature_name)\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1368, in _compute_feature_from_context\n raise KeyError(f\"Unknown BallDrop feature '{feature_name}'.\") from exc\nKeyError: \"Unknown BallDrop feature 'restitution_estimate'.\"\n" + }, + "98fec1b850e24f76bc59e0be4992e509": { + "parameters_hash": "09cf7d9724be516df21be29c1af100da", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-01T02:46:33.058981", + "error": "\"Unknown BallDrop feature 'restitution_estimate'.\"\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1366, in _compute_feature_from_context\n compute = _FEATURE_COMPUTERS[str(feature_name)]\nKeyError: 'restitution_estimate'\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1282, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 879, in simulate_recipe\n feature_dict = compute_problem_specific_features(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 539, in compute_problem_specific_features\n fn(all_signal_simulation, feature_names=requested_names)\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1448, in compute_features\n return {\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1449, in \n feature_name: _compute_feature_from_context(context, feature_name)\n File \"/csem/divr/users/tbe/repo/tsENV/models/simulink/BallDrop/features.py\", line 1368, in _compute_feature_from_context\n raise KeyError(f\"Unknown BallDrop feature '{feature_name}'.\") from exc\nKeyError: \"Unknown BallDrop feature 'restitution_estimate'.\"\n" + }, + "8a2fa4f4d87b45d2990a254a78aaa20c": { + "parameters_hash": "1e0b6a58e2faad407f9bfb3c3cbd9922", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "failed", + "timestamp": "2026-05-01T02:46:36.342435", + "error": "Unable to access ModelWorkspace for model 'simulink_model'\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 297, in apply_ModelWorkspace\n mle.eval(\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/matlabengine.py\", line 71, in __call__\n _stderr, feval=True).result()\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: Invalid Simulink object name: 'simulink_model'.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1282, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1040, in _simulate_case_to_signal_dict\n apply_ModelWorkspace(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 303, in apply_ModelWorkspace\n raise RuntimeError(\nRuntimeError: Unable to access ModelWorkspace for model 'simulink_model'\n" + }, + "eceb4dd37439452499bd677ceab1c1ea": { + "parameters_hash": "ef89459ac7993cc1cc3d3178f9a3050a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-01T02:46:36.835066", + "error": "MATLAB simulation returned no result\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1282, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 922, in run_simulation\n raise RuntimeError(\"MATLAB simulation returned no result\")\nRuntimeError: MATLAB simulation returned no result\n" + }, + "761b44fbee5e464badae4e94c2c8e0be": { + "parameters_hash": "bd94f94fd2672ce364e2d4843dd66565", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "failed", + "timestamp": "2026-05-01T02:46:37.318351", + "error": "MATLAB simulation returned no result\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1282, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 922, in run_simulation\n raise RuntimeError(\"MATLAB simulation returned no result\")\nRuntimeError: MATLAB simulation returned no result\n" + }, + "b02ee6b4abac447fbb757d9fcc70e402": { + "parameters_hash": "ca7cc27af034314228d6d224ceabbf2d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:25.290851", + "end_time_simulation": 9.9975004196167 + }, + "94fd499bdef04535ace846a04b857bf3": { + "parameters_hash": "d4e4c7d3cc25f9d7812468cfc71b78ad", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:23:27.516077", + "end_time_simulation": 9.9975004196167 + }, + "de4bd3ff4118499d9af5e5921aa6c311": { + "parameters_hash": "6c50bdec55f2e4fc34040136f1e8a159", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:29.561180", + "end_time_simulation": 9.9975004196167 + }, + "618d107ba78a4d21a8ede427df093a3a": { + "parameters_hash": "763b089a36df8d26e2432e930d1d1080", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:23:31.766282", + "end_time_simulation": 9.9975004196167 + }, + "b000f6a43e9d4100be9c04c22cdbf536": { + "parameters_hash": "6e831ff5f541c2400eddf5a6a788b963", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:33.822131", + "end_time_simulation": 9.9975004196167 + }, + "f80e701fc3124cb1bf1cf9a5dfb89792": { + "parameters_hash": "5be979d33412c625f2fe6d0ba1739af4", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:23:35.845273", + "end_time_simulation": 9.9975004196167 + }, + "bc78a4f8eb79611383c18e051424dc5f": { + "parameters_hash": "e1731715d402e1e0245b30a492a008e9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:23:38.124557", + "end_time_simulation": 9.9975004196167 + }, + "6824ebaaa1254f499eb2d4ac6ac713eb": { + "parameters_hash": "678286138f07400e0370009202378727", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:40.195664", + "end_time_simulation": 9.9975004196167 + }, + "001063c5dfa248dbbba122ff6743af5f": { + "parameters_hash": "008a1a875535610493e9c18fe12514ec", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:23:46.644196", + "end_time_simulation": 9.9975004196167 + }, + "30965fc8c70344788069076bc18664e7": { + "parameters_hash": "d9c8ec0be3f3e657eda9b02d8230b41a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:48.664968", + "end_time_simulation": 9.9975004196167 + }, + "a6e56514770442afb6af22c381df3a0a": { + "parameters_hash": "95b230d6e9a56e6ff38939b7531573ef", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:23:50.890807", + "end_time_simulation": 9.9975004196167 + }, + "05e552e1f99f43279fbf3a3e4defa6e5": { + "parameters_hash": "0a66a09efea4b9efe7475d035b3c747d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:52.910679", + "end_time_simulation": 9.9975004196167 + }, + "9afc0aea64a5451387e6084d832f3833": { + "parameters_hash": "fea96f2bd94a238c0796569a5f7dd7a8", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:23:55.133859", + "end_time_simulation": 9.9975004196167 + }, + "d9eab0bb66e158e885857119030ed125": { + "parameters_hash": "6f179d33a8f921d84a455dfeae891d25", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:23:57.200694", + "end_time_simulation": 9.9975004196167 + }, + "b75e0120172b4c0ba87c910f93ff072c": { + "parameters_hash": "61493fb3d6a42d1b9b9e497ee3f6a7f1", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:23:59.422939", + "end_time_simulation": 9.9975004196167 + }, + "e6a824386380422eb34461f676432732": { + "parameters_hash": "bbf10de80e8a75804e413786e75f8a6f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:24:01.488145", + "end_time_simulation": 9.9975004196167 + }, + "efeae73360874aefb8d1d01031db168a": { + "parameters_hash": "a38a83f2e3d613aafab279e3f271e272", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:24:03.509970", + "end_time_simulation": 9.9975004196167 + }, + "a1c583110649c9166d2b8cc4ab82eca6": { + "parameters_hash": "88f04e706313e248a820796b502a1348", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:24:05.767853", + "end_time_simulation": 9.9975004196167 + }, + "992e4739fb964b0292c99a2aa7945aa5": { + "parameters_hash": "1d3c5bbe2453d99a6f5acb5e8b7191b7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:24:07.809317", + "end_time_simulation": 9.9975004196167 + }, + "3a5ad8a0c7a84fd3a5baaec0eee7a126": { + "parameters_hash": "4aeb4e670f05f569f0f8133a3cfedd77", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-05T19:10:24.368164", + "end_time_simulation": 9.9975004196167 + }, + "8a79c6b033ed5103a2b8aa51f15c87d5": { + "parameters_hash": "a16634176fe143a740d4f0891973e598", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:10:27.769625", + "end_time_simulation": 9.9975004196167 + }, + "ab2618c21bea42a5b9d58b93173c082c": { + "parameters_hash": "dc6600b666dd665684df177b2470b81f", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-05T19:13:25.085816", + "end_time_simulation": 9.9975004196167 + }, + "844f00b6a1ec4abda2e95dde324ff926": { + "parameters_hash": "0f88a91455c5e7e76bde40ee4c40a457", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:13:29.044328", + "end_time_simulation": 9.9975004196167 + }, + "b5641f9340fa4cd08749657345a1bedf": { + "parameters_hash": "25baa67dc197003efd4ab78cd9496276", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T14:07:55.369729", + "end_time_simulation": 9.9975004196167 + }, + "3b87cbe8124841b59619b88eed80ceeb": { + "parameters_hash": "7b2bec7969bcde646736c85f5a9b5a70", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T14:07:59.378404", + "end_time_simulation": 9.9975004196167 + }, + "7d1b3f1a85c049ff9b8bd7c7a18d4041": { + "parameters_hash": "bb7ee2cccaf39d16c61581ac97abc3b2", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T19:15:50.900507", + "end_time_simulation": 9.9975004196167 + }, + "7e819efac49b9697b79472c289e98675": { + "parameters_hash": "368232ccef11ad40c8c26b6c830a7003", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:09.037142", + "end_time_simulation": 9.9975004196167 + }, + "18ac824559b243958b1ac7b68e9e4b69": { + "parameters_hash": "c98db047a9eef00005a3eac95a689a98", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:12.303162", + "end_time_simulation": 9.9975004196167 + }, + "d5321f7031e24581bbd8dbe94e68f51a": { + "parameters_hash": "7b8ed5013ba3ca8b3e41d6ecdbbef547", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:15.407913", + "end_time_simulation": 9.9975004196167 + }, + "76030b44e8a65f2ebfb5375f1cf2a2c5": { + "parameters_hash": "4c0242f153ded6b9a017f4c1f012a12f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:18.526037", + "end_time_simulation": 9.9975004196167 + }, + "dcbbcb9aaba84f95b6e6c410a573039a": { + "parameters_hash": "c0595eea074c071e3ea0d12bf5cb386f", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T19:16:04.508333", + "end_time_simulation": 9.9975004196167 + }, + "3230f148fbe65ca789e4db09fa311dde": { + "parameters_hash": "5ae4e6914253f8e9029833eecc9c1270", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:16:08.154189", + "end_time_simulation": 9.9975004196167 + }, + "ba8cf6c55f104852811bca0fa4f46393": { + "parameters_hash": "84d332cd5e288f76ff04c153ecdbb662", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T19:16:11.273988", + "end_time_simulation": 9.9975004196167 + }, + "4b6522bc5d255afc898541ca267166f6": { + "parameters_hash": "3f26e10d60dbba9340e0248a6310b9d4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:16:14.299492", + "end_time_simulation": 9.9975004196167 + }, + "12468fab1955463b813e11f74847029b": { + "parameters_hash": "e78defd7189961de44bf71c75071bee7", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T19:16:18.171207", + "end_time_simulation": 9.9975004196167 + }, + "ff68781a5c4b48f29f9f7d4f416d49b5": { + "parameters_hash": "20ceae43b1afd1c5008ca3b6e40bb71c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:16:22.204918", + "end_time_simulation": 9.9975004196167 + }, + "08ab339625f2430992ce5609059a6229": { + "parameters_hash": "3d0d70fae92da0bf5ae5257301b83e57", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T19:16:25.877335", + "end_time_simulation": 9.9975004196167 + }, + "d75a7bf9776b50d48abf72b95a90f906": { + "parameters_hash": "bf5c618e398da6d3224da9cb2b57107e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:16:29.129820", + "end_time_simulation": 9.9975004196167 + }, + "3df1d923d25b44d6826383b65210f7eb": { + "parameters_hash": "ee64f619e246220eace28b0c6a53043a", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T19:16:32.801179", + "end_time_simulation": 9.9975004196167 + }, + "5d1fb1502e984657b8baea94604a2898": { + "parameters_hash": "75d5c6d520b2473fabce0022bf368f95", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:16:35.959748", + "end_time_simulation": 9.9975004196167 + }, + "3044f1762eac4229b425c3e811c01448": { + "parameters_hash": "b6a4fceed9057d9cd76193bb71ac1f25", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:24:55.367411", + "end_time_simulation": 9.9975004196167 + }, + "42e0669504784adea7e79a929f5c5495": { + "parameters_hash": "1c6208a72d86d3824f214c3a12ed1468", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:24:57.555356", + "end_time_simulation": 9.9975004196167 + }, + "e4c9d1c8ce794496bf3cfb6d3e3bf92f": { + "parameters_hash": "e1ad88285b034b66cb8a9ee90a0cc55e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:24:59.607484", + "end_time_simulation": 9.9975004196167 + }, + "831770be078648aebe505cfc3add6f65": { + "parameters_hash": "c54666132e0a75eabfa0b838923499e6", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:25:06.102071", + "end_time_simulation": 9.9975004196167 + }, + "bf0a2f0d66b14e18a99de4aedc052eb9": { + "parameters_hash": "06fe5401557674e82e62b755c642ac45", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:08.170595", + "end_time_simulation": 9.9975004196167 + }, + "9469a80f76084842933ee512968eadc2": { + "parameters_hash": "33737f550d7b5bf32a9ca9761e18e797", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:25:10.391372", + "end_time_simulation": 9.9975004196167 + }, + "653fdc1aab3851a59e4f0c0e792b3458": { + "parameters_hash": "4e9294a7602157528c3f17be7b9e6980", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:12.459897", + "end_time_simulation": 9.9975004196167 + }, + "2711a07092314ea681315ca82e5bc51b": { + "parameters_hash": "047ddfaf94e03e7940e09b1ad70616ab", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:25:14.707285", + "end_time_simulation": 9.9975004196167 + }, + "876c7a0f9b1f5e26b38814efadcc4c8f": { + "parameters_hash": "38b02f8d312f2263e408dad9fb215317", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:16.799982", + "end_time_simulation": 9.9975004196167 + }, + "81cb67a56e3b53dabdc75ab3749d1a65": { + "parameters_hash": "3b5055469dd37e90381e8b7a519b8308", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:25:18.827762", + "end_time_simulation": 9.9975004196167 + }, + "dda83dcb9c9d41999f18ce41a8639bc4": { + "parameters_hash": "fc94670deec89ecc1823e6d57d54b6c0", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:25:21.099354", + "end_time_simulation": 9.9975004196167 + }, + "a8323a08923548f4933e2e96282ebc22": { + "parameters_hash": "e286a89b5472a499ed50fe525496239f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:23.163484", + "end_time_simulation": 9.9975004196167 + }, + "717c312a222c5ba6868ee88f20985de3": { + "parameters_hash": "9497a260af2dc59c87c3ba30dbfea15c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:22.082455", + "end_time_simulation": 9.9975004196167 + }, + "abb9b1157871591bb6e4a58b14b31d66": { + "parameters_hash": "83944f50f4b3205a0363a865af24c7c3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:25.725479", + "end_time_simulation": 9.9975004196167 + }, + "39766784659f583399ef48bbd86c3a85": { + "parameters_hash": "836e1a8f486930cc0d6a751ca0de8b82", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:25:29.758608", + "end_time_simulation": 9.9975004196167 + }, + "90172dd34e645d34a8b308cf9f0238ef": { + "parameters_hash": "cc5999c6fb35fe97e8dfc4b3fd403d85", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:31.802347", + "end_time_simulation": 9.9975004196167 + }, + "569056e49b1940b6ab7ab957d9691b65": { + "parameters_hash": "80666389172edbfd4e48c9c799cc836a", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:25:34.051223", + "end_time_simulation": 9.9975004196167 + }, + "1d78d9f08085498c91d96a2c46abb64c": { + "parameters_hash": "76542e5a1cc2b61bb99a864b220a408a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:36.111076", + "end_time_simulation": 9.9975004196167 + }, + "f609c0f4e64c4394bb7e262fdcbb4be2": { + "parameters_hash": "d05844ce91f519bc272d099da4a89009", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:25:38.366590", + "end_time_simulation": 9.9975004196167 + }, + "938b2ab5dd5a5b0986b029fa817df958": { + "parameters_hash": "1455e1484aea12064e736f3f0e84253c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:40.455396", + "end_time_simulation": 9.9975004196167 + }, + "6c7ac291a99852eea1bd5fb4a619a097": { + "parameters_hash": "ac3bba2bc1090b7838d5e1e47312d9f5", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:25:42.718985", + "end_time_simulation": 9.9975004196167 + }, + "bcfa271de4c751478862d047b6691cc6": { + "parameters_hash": "15bab9c7484f73c98b3c341e76d779d7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:44.791262", + "end_time_simulation": 9.9975004196167 + }, + "3c523b45ab52505d8d9b7c74e5ec21e9": { + "parameters_hash": "2271fe320380cb41f386114ce4c3fdcc", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:25:46.788843", + "end_time_simulation": 9.9975004196167 + }, + "9583b6b3a5c548c0893c647f3350dd5a": { + "parameters_hash": "2f867b9ab82db386c070869d95063d53", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:25:48.998389", + "end_time_simulation": 9.9975004196167 + }, + "0b510fb50d664ad9a319a0f1e817b848": { + "parameters_hash": "1f266c5be6440a4497ad3d575821ede2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:51.096236", + "end_time_simulation": 9.9975004196167 + }, + "a58f39410c8a58289fe81de42778819a": { + "parameters_hash": "fa54d7d0b0be210b7c47e2a8a8a777a3", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:29.122547", + "end_time_simulation": 9.9975004196167 + }, + "5ac911ea878c54e78c1f8d20189c5173": { + "parameters_hash": "ba70c969c4a772f78a2d9da51ad49769", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:32.484029", + "end_time_simulation": 9.9975004196167 + }, + "cf5492d1ceb754e6af4c83df1d1e646f": { + "parameters_hash": "32c83999b4b12fc5e691a21ae0123662", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:25:57.645872", + "end_time_simulation": 9.9975004196167 + }, + "49c6055da90d53c3b9fbc40fe216053b": { + "parameters_hash": "9e2ec336e27c3befdf38ed13c1ca2df1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:25:59.660859", + "end_time_simulation": 9.9975004196167 + }, + "568b3e9ed49658d7abaae622daac8cef": { + "parameters_hash": "6b516d1358c9c9ad27e34c5e0fb2030b", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:26:01.964698", + "end_time_simulation": 9.9975004196167 + }, + "bf16e8da7bc95b958688912361164800": { + "parameters_hash": "47834b692b1279465edb77fe46ddc6b8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:03.988733", + "end_time_simulation": 9.9975004196167 + }, + "2e86956a56055b2aa94c0030af2f2f74": { + "parameters_hash": "55f8d69efb3cee7479c9f93cfebf23d5", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:26:06.263713", + "end_time_simulation": 9.9975004196167 + }, + "a6af683c54fd5c45b1b50f84b9115a3f": { + "parameters_hash": "c30fe335675e1441a69e76f9cafe0ea2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:08.327378", + "end_time_simulation": 9.9975004196167 + }, + "2313b83b6a145fb1832ca2048b375d0c": { + "parameters_hash": "c0293fbff59b29579d1603c66d9a7074", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:26:10.384448", + "end_time_simulation": 9.9975004196167 + }, + "694b67399a894440b95c3a0f0cbaea2c": { + "parameters_hash": "3fb4d82b87979bb32611d331d8005a13", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:26:12.679056", + "end_time_simulation": 9.9975004196167 + }, + "a48651b28d30409ea535272d7d8a6d34": { + "parameters_hash": "8b0aa3d9943f4c633ff480526e78efe7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:14.823125", + "end_time_simulation": 9.9975004196167 + }, + "6ce5356ac9d0595c9afe8deaa22b92f5": { + "parameters_hash": "9ef23db2a4d5fe39d737d0f78fe674e2", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:35.839267", + "end_time_simulation": 9.9975004196167 + }, + "a85037db71c35def8cff51bcd280c2f2": { + "parameters_hash": "effd221c45ae1808dc2487650d1a961a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:39.728784", + "end_time_simulation": 9.9975004196167 + }, + "8aaff1d9f2b553d5ae708caa78a8e6ee": { + "parameters_hash": "2048de04ab0050d0bbc864c7a8f7430c", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:26:21.357885", + "end_time_simulation": 9.9975004196167 + }, + "4ea50cc79cf95e66a287eee00774f13c": { + "parameters_hash": "5d412e823dba6ae864f8786bbc702128", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:23.426483", + "end_time_simulation": 9.9975004196167 + }, + "1efff380b32744068999dbcb131668a0": { + "parameters_hash": "ea78e255868a78a3c0344d19e535c6fa", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:26:25.686262", + "end_time_simulation": 9.9975004196167 + }, + "e622c740da8049a5b64b2587da420840": { + "parameters_hash": "5939949e7282b958c76b711e6168c4de", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:27.778183", + "end_time_simulation": 9.9975004196167 + }, + "197f091fd76c541b99ee8dcbbabc25fe": { + "parameters_hash": "488dedc69f88bb728a39804c4d224cd7", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:26:30.000886", + "end_time_simulation": 9.9975004196167 + }, + "5ceba752d00055649ab609ff12bbfe85": { + "parameters_hash": "5d290bf8302c6117d7af278eb430942b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:32.062580", + "end_time_simulation": 9.9975004196167 + }, + "13377e7a15d95697bf892557c27a39f6": { + "parameters_hash": "5617c8b00a0afc580843dfad00687186", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:26:34.338596", + "end_time_simulation": 9.9975004196167 + }, + "cb05a0c6967853b08984d67601309702": { + "parameters_hash": "0528b823323c7a48b630f2b4730996e2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:36.411002", + "end_time_simulation": 9.9975004196167 + }, + "eca6d692400a55e5b3a16fe4d9caf580": { + "parameters_hash": "8bb3665e75e3feabab36023f1343be4c", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:26:38.458487", + "end_time_simulation": 9.9975004196167 + }, + "6ad833b5696749159b6acff4c3455580": { + "parameters_hash": "8e25ff2713c024639f4a438ca24aae5f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:26:40.675707", + "end_time_simulation": 9.9975004196167 + }, + "0e147c2b22264a96995e40d52b3b2c49": { + "parameters_hash": "dfe717575a0c55be48dad48f0a0a3027", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:42.729597", + "end_time_simulation": 9.9975004196167 + }, + "ea9a51d6f57e4103a6ff75e423e5056c": { + "parameters_hash": "34f385e8e5a6414302a669bb571e4544", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:26:49.292887", + "end_time_simulation": 9.9975004196167 + }, + "4c8df32e5fe556b797f1298c59949be3": { + "parameters_hash": "badc2a6698fa70a51c38ecbc3e3aca9a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:51.317161", + "end_time_simulation": 9.9975004196167 + }, + "7df4a902d4ce5c43914c186e7ead3c66": { + "parameters_hash": "13567246713feeef2961580b6ca7dc12", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:26:53.597790", + "end_time_simulation": 9.9975004196167 + }, + "30f4394e631158a1b847a22f6fac1255": { + "parameters_hash": "8c8f21db05c1d0433857baec9aa890f4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:55.696256", + "end_time_simulation": 9.9975004196167 + }, + "d92defff8af1560982c3892123365961": { + "parameters_hash": "1faf732d47244683c210a12edb0ed7ca", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:26:57.952591", + "end_time_simulation": 9.9975004196167 + }, + "78d36343e1b75872b21961e539178ae1": { + "parameters_hash": "23a71951215cc1c518833f9cf4112c39", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:26:59.953133", + "end_time_simulation": 9.9975004196167 + }, + "a11dcd85a74e4076aca8d9102362a737": { + "parameters_hash": "c9e214cb59b711242e5b94a2bc764b41", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:27:02.241367", + "end_time_simulation": 9.9975004196167 + }, + "999128e792d85110a2fb1f024c2bb3f7": { + "parameters_hash": "c6c4c647aa7af3f00708555372658b1d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:04.328545", + "end_time_simulation": 9.9975004196167 + }, + "07f6245f08985bc3a551b97a931d238e": { + "parameters_hash": "9e6ac4f668624ca7a250cc2de2f46bc5", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:27:06.537025", + "end_time_simulation": 9.9975004196167 + }, + "d481feee6e8a5aaaa8613e008654b7a0": { + "parameters_hash": "ec5c003e6bbfc0b4ae4eeda2a8778b5a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:08.621339", + "end_time_simulation": 9.9975004196167 + }, + "4757e7217e7a56d693494ebf2f740d43": { + "parameters_hash": "a58cb5cd7bb9b9e6ced5502311b0b7fa", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:27:10.610353", + "end_time_simulation": 9.9975004196167 + }, + "c3eb6b6c9abe403faec7dd1a28b4378e": { + "parameters_hash": "b29811a5b1b78bf2be855fe3cae949bb", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:27:12.853404", + "end_time_simulation": 9.9975004196167 + }, + "fba99f461b8b4fdeaed7ed2eccc792ed": { + "parameters_hash": "d8cd65e0d14a11ab6571f848ecfdc8e1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:14.900460", + "end_time_simulation": 9.9975004196167 + }, + "e799ab902f045c1d9c29b64ced889454": { + "parameters_hash": "6e08a6210709db704ec39e8113234e44", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:43.633023", + "end_time_simulation": 9.9975004196167 + }, + "6b87a9d88d635646956ab62ef322ddbb": { + "parameters_hash": "f2a4083fe41e1134eddc494bcfa45196", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:46.789572", + "end_time_simulation": 9.9975004196167 + }, + "ccf1034ad54459debd65aff402728604": { + "parameters_hash": "6e2550ab1675b598fad1d9366529228c", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:27:21.500960", + "end_time_simulation": 9.9975004196167 + }, + "749a4f3ac2ef5ad681c04c487886fdab": { + "parameters_hash": "ed7b417a1dd78e1eb0d6de1d7a4a77ad", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:23.615730", + "end_time_simulation": 9.9975004196167 + }, + "a236c7b60f6d489cad9fdc8ef847a425": { + "parameters_hash": "dcf375a81cfcfec3202906d88dda5f01", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:27:26.000549", + "end_time_simulation": 9.9975004196167 + }, + "120f53b9b2b65fb297e6500840d98704": { + "parameters_hash": "554bfcb8c8cf58edb637bd5c929d5f6d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:28.086953", + "end_time_simulation": 9.9975004196167 + }, + "6dfa0e8eaed44ec6991606d53851c0ad": { + "parameters_hash": "7ad1d40d3ad7cff2f6d131f192a0b59d", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:27:30.356238", + "end_time_simulation": 9.9975004196167 + }, + "3d58ac1000db5451b82984c33fc35b9d": { + "parameters_hash": "353cbf27f27c7b9ab870fb55aa183e56", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:32.462753", + "end_time_simulation": 9.9975004196167 + }, + "5f875cc969b55f368b95aee1a44d01e5": { + "parameters_hash": "22685ffed6e8d27bcaa6c0b78e351417", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:27:34.674338", + "end_time_simulation": 9.9975004196167 + }, + "f719311164db57a895160068decf4b62": { + "parameters_hash": "d15b201c5e54abf5f08166119b924430", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:36.749211", + "end_time_simulation": 9.9975004196167 + }, + "79d335ce3ed85e29bf86d1e1538065e6": { + "parameters_hash": "bd68a77dc8753c6299ab80115403394e", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:27:38.837891", + "end_time_simulation": 9.9975004196167 + }, + "5806802b8340478e9d1ab6d41cfae32f": { + "parameters_hash": "8591b5bfe540bcfe3e8e23da935729d5", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:27:41.112252", + "end_time_simulation": 9.9975004196167 + }, + "f094554d211d4c4fb1711193f9fcb951": { + "parameters_hash": "0e5f3e3f6409be890e93d3609fef9cfd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:43.215855", + "end_time_simulation": 9.9975004196167 + }, + "99f98b55f34e5296803b50dbf5ab893d": { + "parameters_hash": "af58067a9aebed3bf95c03b9cb3be62a", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:50.445063", + "end_time_simulation": 9.9975004196167 + }, + "c0eda41d7e1e5bf0a6470c3308167392": { + "parameters_hash": "4e24a5993b1ca2630937795acdc305a4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:24:54.015596", + "end_time_simulation": 9.9975004196167 + }, + "293ccb493af65394af723a1a23acc7b4": { + "parameters_hash": "5d731b5891207cbe84edd3d7cf040b17", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:27:49.772571", + "end_time_simulation": 9.9975004196167 + }, + "afbf69efc22a53578c3b0ca1cda90eb6": { + "parameters_hash": "6a8b77c65f0d5c77f8ff7848c3d02b88", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:51.884509", + "end_time_simulation": 9.9975004196167 + }, + "a8b3f26fb5f348e5bea777e3d6bc00a0": { + "parameters_hash": "383ab37dfb209e6a1700b2e0de368b39", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:27:54.157911", + "end_time_simulation": 9.9975004196167 + }, + "8f6b1d1dcf66493fad2d43202c865965": { + "parameters_hash": "8000b20ce70f7454a68a471ab244fed8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:27:56.208471", + "end_time_simulation": 9.9975004196167 + }, + "a267997699f95b878bca51ca03b7211c": { + "parameters_hash": "b5860bffb331ed3ca688d05ae8b4df1c", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:27:58.475386", + "end_time_simulation": 9.9975004196167 + }, + "c10ab08e90275b2ba8047c72de0b9caf": { + "parameters_hash": "27e15bf8204ffa376ea510c871610f53", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:00.528285", + "end_time_simulation": 9.9975004196167 + }, + "4373eb3bac935c4e8b80d82e28f50ffe": { + "parameters_hash": "bbe6a229acb51751e042f219741dd051", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:28:02.564528", + "end_time_simulation": 9.9975004196167 + }, + "fb82f39cd2494627a29378b8e4851fb4": { + "parameters_hash": "a0225b3afaea6b7e4795d3256fae794b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:28:04.852462", + "end_time_simulation": 9.9975004196167 + }, + "e8f93df59474426298476ea0a1069174": { + "parameters_hash": "668358a0309036eb05e5fe035334adf0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:06.964058", + "end_time_simulation": 9.9975004196167 + }, + "a7a514d5d4565e85b4f48ad78735021e": { + "parameters_hash": "05b4ad4ca7a6cb1b75aefd36e6c87267", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:24:57.539481", + "end_time_simulation": 9.9975004196167 + }, + "82e926d9c5625e1994cf7a40ad0acfea": { + "parameters_hash": "c09bc91e35cdfe765b2e97a006197821", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:25:00.816830", + "end_time_simulation": 9.9975004196167 + }, + "1b4abf9677a956e4bb90051b4494f4b3": { + "parameters_hash": "d0f642a5f67ba8ba03c2e73321c87bd0", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:28:13.545317", + "end_time_simulation": 9.9975004196167 + }, + "f6ce9817f5725c13bd774915fd22f752": { + "parameters_hash": "e3128e65ae6da1eed7a579b4d25a7045", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:15.654777", + "end_time_simulation": 9.9975004196167 + }, + "3e706177423c5f289117291cc571bc8d": { + "parameters_hash": "d03490834b5ae6e36583fe7b0b6c0cb5", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:28:17.912791", + "end_time_simulation": 9.9975004196167 + }, + "c7329d92814d5a85aac2cbff0b258c41": { + "parameters_hash": "1ab42a0fb7dc5ac9ff900b47d8c7f461", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:19.983440", + "end_time_simulation": 9.9975004196167 + }, + "81867907eb7c5370b8f73d2718d8398d": { + "parameters_hash": "098790278431a07c1c909e2692de1562", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:28:22.297601", + "end_time_simulation": 9.9975004196167 + }, + "24b66e0e0a085d7aa6b93ebf3cb08cfa": { + "parameters_hash": "55a4d2a00a580b802717bc5fa3af0a17", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:24.407890", + "end_time_simulation": 9.9975004196167 + }, + "cd023a4ad4de4316bba0e94b5bf35d8c": { + "parameters_hash": "27dbdcaaa6d7b3190901c22e0f90ce88", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:28:26.637951", + "end_time_simulation": 9.9975004196167 + }, + "206106a281724ad9b5df695293d0b1b9": { + "parameters_hash": "2bb51e1cfdb3bc1d0c843c430faaf3f4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:28.703014", + "end_time_simulation": 9.9975004196167 + }, + "0d96332e3fae53ff8c19e0445a16f304": { + "parameters_hash": "07cb1ec08998ac5324c96499966d920b", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:28:30.775812", + "end_time_simulation": 9.9975004196167 + }, + "303d59b812a942eeac4a05f71df85ff1": { + "parameters_hash": "fec237f2c7761dd4961fab110cf59a36", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:28:33.074655", + "end_time_simulation": 9.9975004196167 + }, + "a5c3ab6d81ea4629894ae78d45592294": { + "parameters_hash": "b6dfa383b9ffd1b15cd838f04e41ec6f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:35.185089", + "end_time_simulation": 9.9975004196167 + }, + "50b90a6df0e25ac480a3af80bb1b21bd": { + "parameters_hash": "a390bd7b013aeba50d900129c8632ed0", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:25:04.167586", + "end_time_simulation": 9.9975004196167 + }, + "26a386fef281514fa7fbc21948c03dca": { + "parameters_hash": "6f7ac29555ebdfc4456c0dcf62ab71a9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:25:07.948120", + "end_time_simulation": 9.9975004196167 + }, + "00f96c534483544fbf5cb3d14e38272f": { + "parameters_hash": "38aa0b9033d2e8106cfcbf264f4825e9", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T23:02:28.882187", + "end_time_simulation": 9.9975004196167 + }, + "d27207501f445f6bb510e7182f86a688": { + "parameters_hash": "974c62873891432182b2dacfbe711fd9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:02:32.649538", + "end_time_simulation": 9.9975004196167 + }, + "9efdb21c07b5589b9b02b808cfd7d78f": { + "parameters_hash": "d7018f58cb2a1b9def853739f1fc7f9d", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:28:46.380236", + "end_time_simulation": 9.9975004196167 + }, + "786dcc247ed7515eb35b7c9d1ee0b476": { + "parameters_hash": "df8e6bafa4ab9de59436b16d1859c574", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:28:48.439604", + "end_time_simulation": 9.9975004196167 + }, + "b9ea19dc7f5a4f948016b9555143f852": { + "parameters_hash": "e27f905eb9495f6f9a6eb9ae06c59bef", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:44:56.660887", + "end_time_simulation": 9.9975004196167 + }, + "643bdaa5daea5733ad7e283dba30f741": { + "parameters_hash": "d311db64c8a0738a89fdd0a13c8710e4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:45:00.975073", + "end_time_simulation": 9.9975004196167 + }, + "fca8a26ed97e46f5a04e19c22db27fa5": { + "parameters_hash": "88879284370adeb6964750d45f07a613", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T18:45:04.917346", + "end_time_simulation": 9.9975004196167 + }, + "b5d723803a4f4f07a269cbf96b10e3de": { + "parameters_hash": "48353e72737bf450e2584cebeaf8c572", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T18:45:08.284484", + "end_time_simulation": 9.9975004196167 + }, + "da6d6445b3604945aaf4475bf4fc42c7": { + "parameters_hash": "9dcb1a663fe67d5237e8c034b5ef4a33", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:45:11.003439", + "end_time_simulation": 9.9975004196167 + }, + "4954a0d8d50b4c0587bf161cf6a0f7b4": { + "parameters_hash": "403a11a8eb944a80314508a2cafea378", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:25:11.477977", + "end_time_simulation": 9.9975004196167 + }, + "27271a9dba095e89b4996ae3777f1372": { + "parameters_hash": "bc58e2b11f93281286378f442803f22e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:25:14.505449", + "end_time_simulation": 9.9975004196167 + }, + "e5c558100e9d465dadac26bb1830e3d5": { + "parameters_hash": "557b33770e7af1f73bf251bcb5ef56ce", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:45:14.181845", + "end_time_simulation": 9.9975004196167 + }, + "fb4a3bd598665f25893f53225e6eb6a6": { + "parameters_hash": "106d8e22968f0c82a8c6950a1934a5ef", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:45:17.493479", + "end_time_simulation": 9.9975004196167 + }, + "d5e208a1a02e48d4ba8d47f9cf10a514": { + "parameters_hash": "0ca42568cddbf4e65e627772ee841644", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T18:45:21.782888", + "end_time_simulation": 9.9975004196167 + }, + "7c4ca3ff4a9341f8a1049cc542344d5c": { + "parameters_hash": "276dc90a18905d4c0bc2c73561be9e02", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:45:25.245074", + "end_time_simulation": 9.9975004196167 + }, + "0f0862eb12d847118678f3f418a5b133": { + "parameters_hash": "076bbb20d83bb845b5fed942de08fa9e", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:45:28.795254", + "end_time_simulation": 9.9975004196167 + }, + "3963cfd01f0a51e1b6f75fa289ff1d35": { + "parameters_hash": "94fcaa225e2d671f809a956dc02148ea", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:45:32.618515", + "end_time_simulation": 9.9975004196167 + }, + "3e28f2d025fd4f229ce62a07f1ed72eb": { + "parameters_hash": "18f315511ca399ede8d2a8bd7ca39601", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:45:37.253942", + "end_time_simulation": 9.9975004196167 + }, + "48b64e9fd8354da0bc49552a546aeced": { + "parameters_hash": "e5c58a48ee7f333858405ee68f0a390f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:45:41.081770", + "end_time_simulation": 9.9975004196167 + }, + "b587da7834684a6a9513f7e7b90e8a9e": { + "parameters_hash": "b38204c1c0625735aba0e9b7d875e92f", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:29:31.762656", + "end_time_simulation": 9.9975004196167 + }, + "a49ddf703986497ea721107abdbda70c": { + "parameters_hash": "5957afd5d4c558ad2d2b7a9cb0fcab0d", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:29:34.033555", + "end_time_simulation": 9.9975004196167 + }, + "c37f1e5b20ea407e9fe3bdef7589e19e": { + "parameters_hash": "bd1192c612b8e303a37a60673021bab9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:29:36.139521", + "end_time_simulation": 9.9975004196167 + }, + "49fd4ebc259940b0bd2709fc346a7cfd": { + "parameters_hash": "95dabc5a4318128f2d8427383c4e1919", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:29:38.414071", + "end_time_simulation": 9.9975004196167 + }, + "85cde74c51544d4888d751d32b4e9ef0": { + "parameters_hash": "8a8e030b83e6b0f628df6a3e3fd1f215", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:29:40.510252", + "end_time_simulation": 9.9975004196167 + }, + "34d91cfcde1b45d6b141ca143c6ea5e8": { + "parameters_hash": "d83176ffdf3547a36ee46068244fb960", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:45:44.542495", + "end_time_simulation": 9.9975004196167 + }, + "511dde78844641f2ad661d8dc0890787": { + "parameters_hash": "c657fe5cef5ee7f8ca00614f7567e783", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:45:48.732756", + "end_time_simulation": 9.9975004196167 + }, + "17a033a89614421b8bb520c3bb286b60": { + "parameters_hash": "08ce3902b9492043d7b112e21f5a3c03", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:29:47.068651", + "end_time_simulation": 9.9975004196167 + }, + "a775cf2b04754935b5d3b2c0f3d6c40d": { + "parameters_hash": "3aea16ef9cb98b406de24ec224ca8b5f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:29:49.146569", + "end_time_simulation": 9.9975004196167 + }, + "919c229adaf44cd68af7629ae8feb882": { + "parameters_hash": "6b8b779f68fdbc2035dac9cc6fe388a9", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:29:51.425915", + "end_time_simulation": 9.9975004196167 + }, + "96553fdefa854f30b03304863919044e": { + "parameters_hash": "3fd56057317ad39374dc5bfc176c4b64", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:29:53.452215", + "end_time_simulation": 9.9975004196167 + }, + "2348ceba66234f5f91acc812e273a3d5": { + "parameters_hash": "417f45bfde6f5b38524778418a5c2df9", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:29:55.685080", + "end_time_simulation": 9.9975004196167 + }, + "60c7d7227d6740599f5c63b9dab2f584": { + "parameters_hash": "b77ce9ffd261e9b2e1cd1228064fe7d2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:29:57.758957", + "end_time_simulation": 9.9975004196167 + }, + "02a90d571817559098d73d6285a817c0": { + "parameters_hash": "e3bcb36d497c3b8f3086da3c37f5ba74", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:30:04.108589", + "end_time_simulation": 9.9975004196167 + }, + "686b8fc5fe3452ca9e924a56249ee369": { + "parameters_hash": "3b3c9b60df1a81b16e780bc79e0ee18f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:25:17.705663", + "end_time_simulation": 9.9975004196167 + }, + "869ecaab7801597d87af479e7440bc45": { + "parameters_hash": "99c95ca68e41b5fac7bf68b70104d9a3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:25:21.139089", + "end_time_simulation": 9.9975004196167 + }, + "a0ca4e2cbf4155f39746b91da71027e6": { + "parameters_hash": "e9d748b3ced10830504cea4c899f3d8b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:30:10.679497", + "end_time_simulation": 9.9975004196167 + }, + "e160b14bc1c55751a056955405acbbe9": { + "parameters_hash": "91e6ef91cab2b5dc134c963b3be3a08a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:12.755970", + "end_time_simulation": 9.9975004196167 + }, + "de0a8239b3b4526fae9d334ca17832bb": { + "parameters_hash": "3a5da17990265ccfe29d7d0aff7b38b9", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:30:15.004381", + "end_time_simulation": 9.9975004196167 + }, + "90c60cd4b1ff5f9282e15a1f0dd56072": { + "parameters_hash": "cb134b4df7ceba14e640baef0c07a157", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:17.111336", + "end_time_simulation": 9.9975004196167 + }, + "854527bb351b50829dde89c49a28aa72": { + "parameters_hash": "2b2976f0c46a91c8de48720eaff5fe07", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:30:19.336053", + "end_time_simulation": 9.9975004196167 + }, + "ea2fd72d2b8057b4b13007da59f3101e": { + "parameters_hash": "796b6ee2d2e88a4200655cf2b626733e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:21.429506", + "end_time_simulation": 9.9975004196167 + }, + "0db2447cdfa559e48582dc2e9b6f7558": { + "parameters_hash": "488f1d5285f322707a4eccf1a9bba94a", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:30:23.720223", + "end_time_simulation": 9.9975004196167 + }, + "7adb560409d25c18a76bfb884ef11c1d": { + "parameters_hash": "90ec7d40a08f3f56ac3611292a7611f5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:25.792522", + "end_time_simulation": 9.9975004196167 + }, + "5b23dea4b1e05cf0bdbbb5fef20c7778": { + "parameters_hash": "338ac51344d86908f3d4bb5da27add46", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T12:39:43.669928", + "end_time_simulation": 9.9975004196167 + }, + "3c3216ccc8ce5b8db589e35abc19bfe8": { + "parameters_hash": "ef250430ec5d695a475a77456f51e840", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T12:39:46.124243", + "end_time_simulation": 9.9975004196167 + }, + "1398b347687256a3a4384f4aa999cf52": { + "parameters_hash": "458e6ea3d33232b15b50fcc457b8dbfd", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:30:32.343446", + "end_time_simulation": 9.9975004196167 + }, + "3307fe4563945397b3c2deb587bfc1b8": { + "parameters_hash": "f8715f1f03d04ad8a3e16c815f751d53", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:34.432625", + "end_time_simulation": 9.9975004196167 + }, + "a4380fadce6c5197815a3216754fd5b4": { + "parameters_hash": "559acfef098b9e93beac99190e0c427a", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:30:36.533134", + "end_time_simulation": 9.9975004196167 + }, + "3fa14ccb8e675dc89c0360344b5cfc4c": { + "parameters_hash": "d2498894e30c06f1a1149bae5109237f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:30:38.787097", + "end_time_simulation": 9.9975004196167 + }, + "4f400f2425ec562a960623051cbb0c5d": { + "parameters_hash": "35bcbb95631f5644e69de2b22b1a52e7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:40.827518", + "end_time_simulation": 9.9975004196167 + }, + "926d3711754f5c1db528a75ef7d5b136": { + "parameters_hash": "aa3a5dc2cc11bd8e715d71400bef3c37", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:30:43.081195", + "end_time_simulation": 9.9975004196167 + }, + "911c01a1996c5615981075a4f29d85c9": { + "parameters_hash": "4e9797a1e39675742a3e3b85b40a45b0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:45.148996", + "end_time_simulation": 9.9975004196167 + }, + "ed3c1a4492c55983a69a4c970b5b0fff": { + "parameters_hash": "31f7fcb1df3abd130e205476bb6f9b2a", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:30:47.468234", + "end_time_simulation": 9.9975004196167 + }, + "7d7960de03bf507ca936df397fa2cba1": { + "parameters_hash": "73638dfe38124c58cf2630cb0718b743", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:49.551870", + "end_time_simulation": 9.9975004196167 + }, + "4210d2506f5252ad83831843fb98b89b": { + "parameters_hash": "7f4c28fcc02cd69cf443db9014db8ad2", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:30:51.859740", + "end_time_simulation": 9.9975004196167 + }, + "c2d7716a947b5121be4818131bba0021": { + "parameters_hash": "117f61f74cbe9f101a2f98c8899a9e8b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:53.981818", + "end_time_simulation": 9.9975004196167 + }, + "e4af54b7f857555497da142e677cd277": { + "parameters_hash": "580369c3e84021a1718afef1c0920379", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:30:56.286242", + "end_time_simulation": 9.9975004196167 + }, + "830ef3f0fa935003803c7ce9a6248878": { + "parameters_hash": "1125c2e325a7ee42572e80c38dc55a33", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:30:58.393167", + "end_time_simulation": 9.9975004196167 + }, + "9310d609d58456a899962deef19c6d4f": { + "parameters_hash": "839e2ca9de4471f278f43d27becc1249", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:31:00.696953", + "end_time_simulation": 9.9975004196167 + }, + "4eae1b023e74552693cebab3bf34874a": { + "parameters_hash": "4a6474b1bf7050e5a2f44aab2285fd04", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:02.782041", + "end_time_simulation": 9.9975004196167 + }, + "7df2e9376cf654358d488e51f581b366": { + "parameters_hash": "8ee58fd5b895962534f77d7c538d6b67", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:31:04.837517", + "end_time_simulation": 9.9975004196167 + }, + "523c9218f52c57068887247f1c7a044a": { + "parameters_hash": "93eac8ecfdcfad4b83c3c05bb8b69549", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T12:39:49.223209", + "end_time_simulation": 9.9975004196167 + }, + "f33eef3b159e5fd2a7ce2072c5f46f10": { + "parameters_hash": "c97af7149b9bc02c78085f87a338aa87", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T12:39:52.378536", + "end_time_simulation": 9.9975004196167 + }, + "972f30e012f55d5fb97587e9a230351c": { + "parameters_hash": "42eaec7e4af9b75dffdfa2fcfdb83e6b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:31:11.560011", + "end_time_simulation": 9.9975004196167 + }, + "c96a04b46b645476a50f16fd142ec1f8": { + "parameters_hash": "08ae791228e0a79c7ed1721eb6a57591", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:13.662979", + "end_time_simulation": 9.9975004196167 + }, + "6514e4c5d16f5fac9810e4b4e0db5d24": { + "parameters_hash": "63f70e3f124cc733cb2ce87a84dece0e", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:31:15.916474", + "end_time_simulation": 9.9975004196167 + }, + "143cced272d1515b9cd99a3d4a487718": { + "parameters_hash": "0c2fd4f9ea097777dd51689b5dbab956", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:17.974782", + "end_time_simulation": 9.9975004196167 + }, + "7e0f85ac769c5ab78201f4cdac386d36": { + "parameters_hash": "ca75e464db9aabb4ec7dd5f7fc14e650", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:31:20.277009", + "end_time_simulation": 9.9975004196167 + }, + "1c2cb974dd5d5d3d9746e81e52664f8d": { + "parameters_hash": "62369822f2e0e1db6af0ac4b21474941", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:22.392872", + "end_time_simulation": 9.9975004196167 + }, + "242699be177953839b3be8c5552afde9": { + "parameters_hash": "79dfe68a5f362e65c5a673f9f9ba0a6e", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:31:24.657090", + "end_time_simulation": 9.9975004196167 + }, + "36c4a8d084d15c9a8b82ea18c288b196": { + "parameters_hash": "bc26217e05e254401e40db88d0f7ec83", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:26.789200", + "end_time_simulation": 9.9975004196167 + }, + "33508fdf1d3d5b62a0f61a2ab5200441": { + "parameters_hash": "7927f0e02ed2fe57d1885a23b2a4de66", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:31:29.004671", + "end_time_simulation": 9.9975004196167 + }, + "701e87397be55c36bf795804cac2fa82": { + "parameters_hash": "cb097a8cfa809c98f01f26887c1d1d7d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:31.099963", + "end_time_simulation": 9.9975004196167 + }, + "bb0abe5b86f652bf819eb1c65190c4c7": { + "parameters_hash": "64c3874fe139e7f82cea10a84ad55682", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:31:33.124821", + "end_time_simulation": 9.9975004196167 + }, + "f4c8f20b5816534db655784eb8d032f6": { + "parameters_hash": "e36e5fb8caecbbb6fa7e910f2a30d7f8", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:31:35.373892", + "end_time_simulation": 9.9975004196167 + }, + "2a87e3379e355d448b5ec4da64b93950": { + "parameters_hash": "c77c3b08566506b980711f01ad465ed8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:37.460128", + "end_time_simulation": 9.9975004196167 + }, + "299b1d05446c56c085e14c21e239a7da": { + "parameters_hash": "e2c9c8caeb332a260528d855192490e3", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:31:39.723970", + "end_time_simulation": 9.9975004196167 + }, + "01652097fe235d158c476b7d629f1715": { + "parameters_hash": "24d070b6baa0ac026ca1e80c076314d2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:41.740945", + "end_time_simulation": 9.9975004196167 + }, + "7da3ecb6fce8564ab329efd4235c9f65": { + "parameters_hash": "94233207d7157b44bf58584bf409508e", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T14:12:38.719084", + "end_time_simulation": 9.9975004196167 + }, + "3bfdbd5d7eb15107a3f092eae73ff463": { + "parameters_hash": "ec258fa1eab9ebd795ca1bba43f25237", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T14:12:42.344894", + "end_time_simulation": 9.9975004196167 + }, + "b70e19c7713d570fb886ee36e75a7fd4": { + "parameters_hash": "0d3315ab1de852fb002dce5dde3138eb", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:31:48.391358", + "end_time_simulation": 9.9975004196167 + }, + "644bcf8598935e908eb28c45fd550a6b": { + "parameters_hash": "0170c6037bfca174156702989c264fb2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:50.472087", + "end_time_simulation": 9.9975004196167 + }, + "215fe82e5df953bd86e78e19ec8b88d0": { + "parameters_hash": "7b9796d2ddef7160b26ddd78fd2f49e9", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T12:39:55.889971", + "end_time_simulation": 9.9975004196167 + }, + "4f7cd2cf633156c08c9280e3b4bfc3ca": { + "parameters_hash": "b3126b813c5123db35adc69b90319a6e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T12:39:58.741385", + "end_time_simulation": 9.9975004196167 + }, + "c8baff863e805d089948e883182b8d93": { + "parameters_hash": "326d77c44cb90667e86e8bd4ca6ccae3", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:31:57.114828", + "end_time_simulation": 9.9975004196167 + }, + "1b9711f003d452c3adfe632996ba5e7a": { + "parameters_hash": "578595111ba82218c86fad4e581ba886", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:31:59.205595", + "end_time_simulation": 9.9975004196167 + }, + "c5e10373855753ac8c8bc93fb52dc00a": { + "parameters_hash": "8f81b180bb84378e5a55b464860078c0", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:32:01.300689", + "end_time_simulation": 9.9975004196167 + }, + "e5078858b66a5abf9e0e3a61de9b5d60": { + "parameters_hash": "d1d18319b3f00ff4cddbd1385f7614e6", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:32:03.579023", + "end_time_simulation": 9.9975004196167 + }, + "79a47d5e9ef75060947c1e1ed7a38e8d": { + "parameters_hash": "dfade4222dbf1af68426c50b9886dcc1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:05.673904", + "end_time_simulation": 9.9975004196167 + }, + "972f4e6a45e75441880c08a3c7cd53ae": { + "parameters_hash": "9a4643019291f2b59e71a35da3e428be", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:32:07.969300", + "end_time_simulation": 9.9975004196167 + }, + "0aa940a907b4538b90640068b6afb65c": { + "parameters_hash": "0eb119b7f90609d558455612a28e9cac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:10.066628", + "end_time_simulation": 9.9975004196167 + }, + "908124c60e0157179286cc9530a4028f": { + "parameters_hash": "a2ce27a4af480a84efba3aa599053be8", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:14:42.283701", + "end_time_simulation": 9.9975004196167 + }, + "6d3a17f5b8f950ababf338b039dbc9d6": { + "parameters_hash": "658c43d47a82499fd05bc0b7e3cb8560", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:14:46.422956", + "end_time_simulation": 9.9975004196167 + }, + "f76551f826f95ae990a1e3e05924c613": { + "parameters_hash": "7a7c5a3c6b8565f3eac397789e0b6e78", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:32:16.669083", + "end_time_simulation": 9.9975004196167 + }, + "24d5a60456b752d3877114f73ff7e9c6": { + "parameters_hash": "06f56b5b971c45a896312bf91b55b77a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:18.759464", + "end_time_simulation": 9.9975004196167 + }, + "5cb3299f0dc0582abdeb6a180123d711": { + "parameters_hash": "cd4ef198ec66c3d31baf82318debc64e", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:14:50.935444", + "end_time_simulation": 9.9975004196167 + }, + "63a7e2f0dddb58078c52818bee66eb05": { + "parameters_hash": "44bebe5e9cdd625a6da77b97aa1cb5bb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:14:53.810111", + "end_time_simulation": 9.9975004196167 + }, + "b87f2159601f5ffdaa1a87d17496a316": { + "parameters_hash": "2c56acb4d610622648d44c800d3cb62d", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:32:29.551360", + "end_time_simulation": 9.9975004196167 + }, + "b49cf1383034516ca6e7de8f55e6afe6": { + "parameters_hash": "440a55621e8cc658a56d02a4ba396f46", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:32:36.168549", + "end_time_simulation": 9.9975004196167 + }, + "03b8bb13840f56aaa042d9233e594182": { + "parameters_hash": "72dc20b034fa89cf70eaef019c0c1a9e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:38.213848", + "end_time_simulation": 9.9975004196167 + }, + "eb87de6e35ba5ef3a44b373e503de244": { + "parameters_hash": "2e7ba38be72d7ffe11fa766beff0bf39", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:32:40.457838", + "end_time_simulation": 9.9975004196167 + }, + "b7aab5d6ddeb5393964f2516b8a55598": { + "parameters_hash": "9690b16d0553eca4b4dd0495526071d7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:42.549280", + "end_time_simulation": 9.9975004196167 + }, + "33477f0d3e57542182e18d90468b6e3c": { + "parameters_hash": "2386e07b916b0b09fdc05b99b6bf125b", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:32:44.777966", + "end_time_simulation": 9.9975004196167 + }, + "edc6cea995935fb1be150f41fd647b06": { + "parameters_hash": "e28dfd6123e65a9553e2d793d6fb0296", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:46.877390", + "end_time_simulation": 9.9975004196167 + }, + "ffc379dfbab0528c900ac440866a90e2": { + "parameters_hash": "b7e2f83fa7c510c85f71e143cf9e5406", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:32:49.141327", + "end_time_simulation": 9.9975004196167 + }, + "cdd8388d72935366a2eb7d7ac0567cff": { + "parameters_hash": "84a6885a789e3d9702a7ff9fab05abe3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:51.225944", + "end_time_simulation": 9.9975004196167 + }, + "63b4f438163057f9b67ff2a42d6427c6": { + "parameters_hash": "2a7ceb16692073467365faa0c1cbd44a", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:32:53.251502", + "end_time_simulation": 9.9975004196167 + }, + "8f251b7ab3515230926f69d9fcd57dd4": { + "parameters_hash": "45e01075619fadbd7a23034d6ffdc43c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:32:55.478107", + "end_time_simulation": 9.9975004196167 + }, + "995c1a2bcc965ae6b8fd3abffc107a38": { + "parameters_hash": "cf8ec2a84c7e7b6b2ab9f47016a3b916", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:32:57.534546", + "end_time_simulation": 9.9975004196167 + }, + "7f19fa1121e35a18b693a5919a5e9e3d": { + "parameters_hash": "5e5832720ee31a070d9527801fa12a3d", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:32:59.775254", + "end_time_simulation": 9.9975004196167 + }, + "fc4c4751b13e5beca3709f3886396e4b": { + "parameters_hash": "55c97700474ed98646bf8d95d091fab6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:33:01.880441", + "end_time_simulation": 9.9975004196167 + }, + "56329b0d596f5b4daa3db101e5a634e6": { + "parameters_hash": "b67fddc78fbf439909b7f85258550959", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:33:04.156906", + "end_time_simulation": 9.9975004196167 + }, + "a7278b34fc94532ba2439dcdaa2213f3": { + "parameters_hash": "8e39e21a4f96afc520cd8401a434ec82", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:33:06.190448", + "end_time_simulation": 9.9975004196167 + }, + "f73e9eeb29be5e07b8c13705ff2ab2a6": { + "parameters_hash": "77dbaedb84822756a608e441ea18ff96", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:33:08.499876", + "end_time_simulation": 9.9975004196167 + }, + "800e64109f5652ca9d0f4038cf689b89": { + "parameters_hash": "c4633e7c8bc342e125114a14143790ed", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:33:10.601281", + "end_time_simulation": 9.9975004196167 + }, + "d72e4ead5b465703b756c36d29213e3f": { + "parameters_hash": "89c7778dfb923796d6e6ec8ef5758779", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:33:12.877301", + "end_time_simulation": 9.9975004196167 + }, + "01a618f687a95bb6867831251bba927b": { + "parameters_hash": "884869abc0a73fcc7cec0e30a5237d75", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:33:15.002168", + "end_time_simulation": 9.9975004196167 + }, + "873c2abc53235d5eb408430e7313c0b7": { + "parameters_hash": "044d90f79212984839c168124645adf7", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T12:40:02.091310", + "end_time_simulation": 9.9975004196167 + }, + "00ddbb5e94655de890433946e4d0f398": { + "parameters_hash": "ff4eb64793fa6c21897b1ce7422ca660", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T12:40:05.824216", + "end_time_simulation": 9.9975004196167 + }, + "c5333739516d5fbfb8aa22aaf11a7cc9": { + "parameters_hash": "4244ddfb23d3513ea6438f4ab22dfb58", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:33:21.620846", + "end_time_simulation": 9.9975004196167 + }, + "dff4efa541c35605b59006ff628c641a": { + "parameters_hash": "eb47deb247568a3f13c1a48997a7ccf6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:33:23.715594", + "end_time_simulation": 9.9975004196167 + }, + "61e04dcf1460548aa8646c442a0784f1": { + "parameters_hash": "14a3887414efa81b4e4984c37902203b", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T18:10:10.550301", + "end_time_simulation": 9.9975004196167 + }, + "5f0c1ddf46645f38936785be89afd781": { + "parameters_hash": "13d3c1c2a136d410923b65e7ba4af872", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:25:24.938426", + "end_time_simulation": 9.9975004196167 + }, + "d482e337fdc45cc89bc72fc2170ac3b0": { + "parameters_hash": "03d974d1d6e0c75ffef7f5b5ab55a1e1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:25:28.132332", + "end_time_simulation": 9.9975004196167 + }, + "eaa32e6c103f5e2cba5b8116bcb2263d": { + "parameters_hash": "5f307605e397c4e6d59a65888e7db42c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T18:10:16.967461", + "end_time_simulation": 9.9975004196167 + }, + "f42cafbb5b315f659db0e7e81a8bdb45": { + "parameters_hash": "10432350d0f8ff36cd5141c4e4a0c4b5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:10:20.520134", + "end_time_simulation": 9.9975004196167 + }, + "9035a21f0a2e5bea99434b1e9b5e966f": { + "parameters_hash": "804a2b7baee148335b40db593c10e920", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:10:24.913639", + "end_time_simulation": 9.9975004196167 + }, + "fc07cde3cef8525480ee4ca871fb017d": { + "parameters_hash": "a28bad2a8da087ff3e79676a3bb78bbc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:10:28.133710", + "end_time_simulation": 9.9975004196167 + }, + "672cebaad55d55febcee5de9850a490c": { + "parameters_hash": "85474877919a0e7f70dbc708f48a43b0", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T18:10:30.893693", + "end_time_simulation": 9.9975004196167 + }, + "c120ad11aeb65ec08aa280f32e82990b": { + "parameters_hash": "8ee71345cd9d66f06c0d7ac69747ef8b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:10:34.120951", + "end_time_simulation": 9.9975004196167 + }, + "746b0d955ae259e79ed354f101e63d77": { + "parameters_hash": "04988d58c1ff0245168df31282ff385d", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:10:37.599021", + "end_time_simulation": 9.9975004196167 + }, + "fda8f8ca3d905389bb126773bdecec8f": { + "parameters_hash": "65f557584b06b601396dfa5a6d32636e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:10:41.181954", + "end_time_simulation": 9.9975004196167 + }, + "34574575af47569e8859d01ec8078234": { + "parameters_hash": "b40c434de3e35557c8d3bf3db99a4066", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:33:54.032905", + "end_time_simulation": 9.9975004196167 + }, + "f5ff4fd3cd685dd4baee7c5bd21e90bb": { + "parameters_hash": "ac9fe672bdb2af32aca753b90dffd0f1", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T18:10:44.458317", + "end_time_simulation": 9.9975004196167 + }, + "b5ab195f5a0a59029732a38693bac36f": { + "parameters_hash": "0a61ac69b41508a46f8cb1e8e86d35ff", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:10:47.640676", + "end_time_simulation": 9.9975004196167 + }, + "d5e619d9d7fb52898940b8140f653c12": { + "parameters_hash": "a0e3858acd774f53864295bc32fb89ca", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:10:51.027710", + "end_time_simulation": 9.9975004196167 + }, + "6b6348c2313355ee8813bf17f88375ab": { + "parameters_hash": "694d59f0fbde763bbccfdf7262f0a07f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:10:54.068312", + "end_time_simulation": 9.9975004196167 + }, + "9daf764b6051540e98b175ec43c551c4": { + "parameters_hash": "104c2d85c6ad7c82eb0d205855155b1b", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T18:10:57.582093", + "end_time_simulation": 9.9975004196167 + }, + "5ed261f9b859543db1bbc43a0e484e88": { + "parameters_hash": "02514731a68f154a69f172e2a4d1dbbc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:00.851909", + "end_time_simulation": 9.9975004196167 + }, + "ca1727e8ff895ff497d01835818489e8": { + "parameters_hash": "db1b00130d23124ded6b33056368abf6", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:11:04.437320", + "end_time_simulation": 9.9975004196167 + }, + "b8037d9bbc155a85a314f03b8e5ead59": { + "parameters_hash": "339c992972f3c8fe75997c4e52ab2e2a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:07.088627", + "end_time_simulation": 9.9975004196167 + }, + "55cd4372b9205e238624bf60157b1250": { + "parameters_hash": "ae7ad5d062eeca72240280476370ae0b", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:11:10.585458", + "end_time_simulation": 9.9975004196167 + }, + "a47f9fc9ac8d5ba7bb003fbffc8b1b50": { + "parameters_hash": "bfb2c455858c263715da87217cd29d4c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:14.313450", + "end_time_simulation": 9.9975004196167 + }, + "7fa479e6019a5f0799d0f9aec031911d": { + "parameters_hash": "6bb9a32a46580d245e6f605b62d8f2d1", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T18:11:17.766205", + "end_time_simulation": 9.9975004196167 + }, + "09efc5429b8656cf88941eaf685b35d8": { + "parameters_hash": "dd20434e4f46f56c7165a1208e34879b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T18:11:21.447733", + "end_time_simulation": 9.9975004196167 + }, + "850f7c0ddb115f058c442c76ae15e3f9": { + "parameters_hash": "306c9793f4bd430cc60130719fa97faf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:24.902732", + "end_time_simulation": 9.9975004196167 + }, + "37f1cd546bd25045ae1d59fcc25bb7a9": { + "parameters_hash": "2eb3078297045d663e0f4f65d81592d0", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:11:28.310555", + "end_time_simulation": 9.9975004196167 + }, + "1c4f5bd0db71585589652796e23b59e0": { + "parameters_hash": "4ed1ffcbe53c885723c898675ef9734e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:31.818056", + "end_time_simulation": 9.9975004196167 + }, + "0aec15477eef5733869f01588e4b4a10": { + "parameters_hash": "432c1016af8376de1f505807c53a6c4e", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T18:11:35.888337", + "end_time_simulation": 9.9975004196167 + }, + "730695c984475f8d8bc5a8d12b39ae44": { + "parameters_hash": "59736633152948478662b3432626556e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:38.983679", + "end_time_simulation": 9.9975004196167 + }, + "4ab30a18bc29570cb06b030e0a43f146": { + "parameters_hash": "68eceffe300d68510e5137b58d476221", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T18:11:42.621531", + "end_time_simulation": 9.9975004196167 + }, + "08475b87e5075d789d5ac6a6114cf8c7": { + "parameters_hash": "84ec46837888d37ab42e62ee9981c28e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:46.705619", + "end_time_simulation": 9.9975004196167 + }, + "f0f496acd2f350a7868d34de7c1dbd8f": { + "parameters_hash": "8d9d564fa4932eed84ba5222100a4bea", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:11:50.506356", + "end_time_simulation": 9.9975004196167 + }, + "b1be25c64d4157adaef6ff798d005be4": { + "parameters_hash": "561a5ab547bbccc5b37ead44cc66d429", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:53.828327", + "end_time_simulation": 9.9975004196167 + }, + "22e7aed96902591f976f0c03805537f0": { + "parameters_hash": "07512c30945f233ff2a157c7f7cc0d86", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:11:56.905624", + "end_time_simulation": 9.9975004196167 + }, + "8037061f29675b25a8426c7c6692b9b9": { + "parameters_hash": "aa8d416bd2c1da5a9f8cd2aa2a87ece1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:11:59.613582", + "end_time_simulation": 9.9975004196167 + }, + "7109214388a753679ea3347d46ee75f5": { + "parameters_hash": "abe77eae50b08312ecaf46197fba9807", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:34:55.548178", + "end_time_simulation": 9.9975004196167 + }, + "fbe454d4b3625caaaa07aaed18d16268": { + "parameters_hash": "664e12e057bc5745ba8b6a0b17907b2e", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:34:57.814865", + "end_time_simulation": 9.9975004196167 + }, + "ea17d873e2cf5d658b3b00158ae43c95": { + "parameters_hash": "a13653217d9bb930d4169669025a413e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:34:59.905577", + "end_time_simulation": 9.9975004196167 + }, + "9c445001c2f75ddb9bd1d47993a664f2": { + "parameters_hash": "c3fc753c70d195ba992e129fd1bfe55a", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:35:02.208030", + "end_time_simulation": 9.9975004196167 + }, + "33d64a98a2ab59909873d826d29d9296": { + "parameters_hash": "512dfc844b1831ab9d094aa7a2ea841d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:35:04.298205", + "end_time_simulation": 9.9975004196167 + }, + "74558dec52105edfb945508fea345732": { + "parameters_hash": "d75932dfc68110c7928d725d1a6f2b8f", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:35:06.564937", + "end_time_simulation": 9.9975004196167 + }, + "2fd8f2c2147e58308465d62e025d19a5": { + "parameters_hash": "50339848ee6e9fb8c7894fd79ddc0c31", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:35:08.682067", + "end_time_simulation": 9.9975004196167 + }, + "93651e54ac445eba83c9f71cd64421a9": { + "parameters_hash": "7bb74baa7f937506c2dafd80ce10b72e", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:35:10.949113", + "end_time_simulation": 9.9975004196167 + }, + "7b1edb225dea5360a6a56ccec46b1dcd": { + "parameters_hash": "69f29b2457896cb1be6871419bb7aaba", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:35:13.017707", + "end_time_simulation": 9.9975004196167 + }, + "36aee79d125e52d18d0608ce1873c6b4": { + "parameters_hash": "2fa0662ab14d66a321fdab622c17c590", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:35:15.311320", + "end_time_simulation": 9.9975004196167 + }, + "d70a159da396546cac6515377538ad2f": { + "parameters_hash": "d38b3f2482fb21262679ae743e22e350", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:35:17.456616", + "end_time_simulation": 9.9975004196167 + }, + "1dcbb406a6365e2f86382f83d5a8a508": { + "parameters_hash": "156430334bacc24f18dce6562ddb8dd0", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:35:24.047988", + "end_time_simulation": 9.9975004196167 + }, + "835dc6942a055feea8d0e0eddbdcd0ea": { + "parameters_hash": "0a4f15886fa0d66df90fadd8873714ff", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:35:26.135552", + "end_time_simulation": 9.9975004196167 + }, + "ea6801d9099c5277a09e86a4b622c931": { + "parameters_hash": "7038b1926c27b4414dcfb232e8179603", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T18:12:02.517890", + "end_time_simulation": 9.9975004196167 + }, + "47027d922be35c2f9a8bfcbd3d341b33": { + "parameters_hash": "c56c1447e31af1244ad611a1908b7a76", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T18:12:06.359886", + "end_time_simulation": 9.9975004196167 + }, + "60edcb546d955607ac58d96394fe9707": { + "parameters_hash": "28babefcffcea7004e7ac3525265fdb9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:12:09.612075", + "end_time_simulation": 9.9975004196167 + }, + "122b853b34665683b24281d8f2060407": { + "parameters_hash": "30a5af9d6e6b5297d6ef05407f1fd433", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:12:12.789281", + "end_time_simulation": 9.9975004196167 + }, + "beb11279b14f59008ecbb4fa95c9097b": { + "parameters_hash": "af4b5afeec877f92be99ea648663b6a5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:12:15.875066", + "end_time_simulation": 9.9975004196167 + }, + "7c4e6a1c9a7c5ddeacffdbf9ff1452fd": { + "parameters_hash": "b53a7638f7841757c0f63d09e81183a4", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T18:12:20.192383", + "end_time_simulation": 9.9975004196167 + }, + "10efb79f2e515c84afa17e4477de1943": { + "parameters_hash": "556399c38c97e3102f34e5d49da67f77", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:12:23.759112", + "end_time_simulation": 9.9975004196167 + }, + "83888e22c96f5b1982d570283bc6ec1c": { + "parameters_hash": "e71a95e47430b116457ba2882cfb0bd8", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:12:27.129424", + "end_time_simulation": 9.9975004196167 + }, + "2c3a08faa4a05458866b1ff32622a493": { + "parameters_hash": "4054d3b7726dac1180c51637e30c7d08", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:12:30.687301", + "end_time_simulation": 9.9975004196167 + }, + "9a9da2ff03535172ad9cc0f39d360514": { + "parameters_hash": "4b720372d9cfde8622522a76dce42884", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:12:34.359362", + "end_time_simulation": 9.9975004196167 + }, + "c4338f3794c25bb2b7a752a8c045afc9": { + "parameters_hash": "8546ad617caa96c3d3593711aac6ee7a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:12:37.597790", + "end_time_simulation": 9.9975004196167 + }, + "47f84b9f0967540a98380bd737333252": { + "parameters_hash": "9860761527be368b6e6e5315b3fb7a8d", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:35:56.556498", + "end_time_simulation": 9.9975004196167 + }, + "fbee8e4e80435cb08b5438b33fb62e36": { + "parameters_hash": "2bec3dbcf5c2516b470d8ed570fe42b1", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:35:58.826327", + "end_time_simulation": 9.9975004196167 + }, + "2da7afda8f5c577e9cca5c8d4358eb34": { + "parameters_hash": "9a9d3548a2ad89f44b8697260e2f2740", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:00.912639", + "end_time_simulation": 9.9975004196167 + }, + "b241861376b15d9a894e5ef60137a618": { + "parameters_hash": "58827de69ad895481ed28f4b0aead479", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:36:03.212986", + "end_time_simulation": 9.9975004196167 + }, + "663ac7d86f85580996c4073b4a896e20": { + "parameters_hash": "34593b18a268667ec89b4c6cacced590", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:05.329042", + "end_time_simulation": 9.9975004196167 + }, + "0180b40867d95f08ad5438d11c5427af": { + "parameters_hash": "efbb7cad0a2d30b3fc01bd40280997f1", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:36:07.604631", + "end_time_simulation": 9.9975004196167 + }, + "f7fe578e1c0e5cb1967bcd7a3deed889": { + "parameters_hash": "f70f54daab02ce1133affc8c1ffdfe68", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:09.761580", + "end_time_simulation": 9.9975004196167 + }, + "6228a7c919375242b6065fac22112b1e": { + "parameters_hash": "2bd1aab36ab5b595de55d70fad4b0509", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:36:12.056337", + "end_time_simulation": 9.9975004196167 + }, + "8d404b6773a45cc9ada11af7abf3e313": { + "parameters_hash": "86850f5db5ab47f5c09109a37b5c15ad", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:14.147561", + "end_time_simulation": 9.9975004196167 + }, + "d47f56abc03d5b9ca55ff2f0f9d0e1d3": { + "parameters_hash": "eebd8ced7079e7cce992af4756e4a1df", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:36:16.440744", + "end_time_simulation": 9.9975004196167 + }, + "1b9f145668945a8fbfa43f0d796c92c0": { + "parameters_hash": "5ed1c068abc301b5e22386cd72bc6c90", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:18.498952", + "end_time_simulation": 9.9975004196167 + }, + "06c7aeb4badb5ca085c44da26f39cfa5": { + "parameters_hash": "9e37b30901ff0d2cafb29885c1d82916", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:36:20.756853", + "end_time_simulation": 9.9975004196167 + }, + "f876719ea1e258e29707d24f570e914c": { + "parameters_hash": "b3fb4f0572272ee865353f1bf86ac317", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:22.913694", + "end_time_simulation": 9.9975004196167 + }, + "13aa5e92152350df81b8434d5d62d0b9": { + "parameters_hash": "1257decc2a98e7b71a279358c5bc1abc", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T18:07:51.216988", + "end_time_simulation": 9.9975004196167 + }, + "ccd74be81423587e9bcca01485748c5e": { + "parameters_hash": "b0b28390c05459f7ca8f17370d6375f7", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T18:07:57.739806", + "end_time_simulation": 9.9975004196167 + }, + "3ccf19df68935a4bbc182728ed770842": { + "parameters_hash": "db1344362c6844f9dba5070620a6ff20", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:08:02.462289", + "end_time_simulation": 9.9975004196167 + }, + "3f68e497f69858feb62a694a12e56efb": { + "parameters_hash": "55656ff08f9b229be681f06f4026ed73", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T18:08:06.461942", + "end_time_simulation": 9.9975004196167 + }, + "57cbe50f881850a7a3f5318506b95de6": { + "parameters_hash": "792b3dc2cb5cf92d57e156d7d8eb1456", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:08:10.097708", + "end_time_simulation": 9.9975004196167 + }, + "87c737a025f251dd9ae3281be3c7d175": { + "parameters_hash": "5944b3fe2a6ecf01f482148db0a344c1", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-02T18:08:14.328809", + "end_time_simulation": 9.9975004196167 + }, + "094a798c298c5fa2ab1f4c0a09b0f014": { + "parameters_hash": "d93fa8b5d981e62b2e2d3bf211f92ddc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:08:17.687285", + "end_time_simulation": 9.9975004196167 + }, + "2e5637bb95b7578eb16bc565e01b1da3": { + "parameters_hash": "bbf56a93a8c706b2f0a889681f281d09", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-02T18:08:21.349761", + "end_time_simulation": 9.9975004196167 + }, + "6d7404e2f7345012b65c9c571ad68c92": { + "parameters_hash": "048942496a0f0ed27fe59897f9bc3856", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:08:23.846731", + "end_time_simulation": 9.9975004196167 + }, + "8abfef34141c50eba70ddbf203145f64": { + "parameters_hash": "3c164763563f53d26b92871cd26f0481", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-02T18:08:27.103637", + "end_time_simulation": 9.9975004196167 + }, + "924677ba27d35e1298690b29b00b70a9": { + "parameters_hash": "025f79b41c7c51cd9c885549a7439d4e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T18:08:30.313217", + "end_time_simulation": 9.9975004196167 + }, + "ba2e3cbfe88c52748c6d2c81c7a04600": { + "parameters_hash": "9bec6d5ac5e767ece2dddf6301cbb7e3", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:36:48.893669", + "end_time_simulation": 9.9975004196167 + }, + "77968ae18bca566cbcffba3830a3fece": { + "parameters_hash": "1556a1f828847ba49d9ade99e0c2de54", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:36:51.207034", + "end_time_simulation": 9.9975004196167 + }, + "e0be924591be523db964016177a605a9": { + "parameters_hash": "11dfcf5ea779e50974ee80d19b763e95", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:53.294481", + "end_time_simulation": 9.9975004196167 + }, + "18e695472b2d52058cf6deee57c49cf8": { + "parameters_hash": "619e92593025d6e33838d323838e183b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:36:55.566259", + "end_time_simulation": 9.9975004196167 + }, + "cd752b36da165351a677c45786835918": { + "parameters_hash": "1692cb009ab6037f1f6d65b69954bd40", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:36:57.690109", + "end_time_simulation": 9.9975004196167 + }, + "e195837b1041539c86a3733e4391198b": { + "parameters_hash": "b382ce19871ef185790c086c584cd601", + "run_type": "intervention", + "class_internal": "drag_coeff", + "class_agent_facing_name": "drag_coeff", + "status": "success", + "timestamp": "2026-05-01T12:36:59.997025", + "end_time_simulation": 9.9975004196167 + }, + "d0943d6408545165a9d5b9c7eb2155b4": { + "parameters_hash": "af4d99e1e086fc717f0117518abe6131", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:37:02.133087", + "end_time_simulation": 9.9975004196167 + }, + "2f63aec58cd05535ad7d8d5044268c18": { + "parameters_hash": "c6874aebe44873711f5897d513d8d246", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:37:04.443261", + "end_time_simulation": 9.9975004196167 + }, + "89aa52be4f2456c4828d4947d260b8f2": { + "parameters_hash": "d23086421ba8dea0ebab73bf73021da9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:37:06.554592", + "end_time_simulation": 9.9975004196167 + }, + "e57e705e527d514db713ff28280d11fd": { + "parameters_hash": "bbeec99eec6011df3dd9567215e34028", + "run_type": "intervention", + "class_internal": "restitution", + "class_agent_facing_name": "restitution", + "status": "success", + "timestamp": "2026-05-01T12:37:08.876536", + "end_time_simulation": 9.9975004196167 + }, + "1b1b4fd87f4152a9b7632b635727e774": { + "parameters_hash": "1342ba5162887b5fca9ccbf2beff4ec0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:37:10.966136", + "end_time_simulation": 9.9975004196167 + }, + "1f176712882c5ce69f5f87c4473b6694": { + "parameters_hash": "39211469f82bdc723cce0f4bb49d586d", + "run_type": "intervention", + "class_internal": "gravity", + "class_agent_facing_name": "gravity", + "status": "success", + "timestamp": "2026-05-01T12:37:17.658594", + "end_time_simulation": 9.9975004196167 + }, + "14deceb4aedd5c938ce6e21acf45cf09": { + "parameters_hash": "16b5e9dbed5cee722b063098aca6750a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:37:19.762805", + "end_time_simulation": 9.9975004196167 + } +} \ No newline at end of file diff --git a/questions/BallDrop/noise_adder.py b/questions/BallDrop/noise_adder.py new file mode 100644 index 0000000000000000000000000000000000000000..bdaf129246496d352a6ebb49f32c63d31596ff16 --- /dev/null +++ b/questions/BallDrop/noise_adder.py @@ -0,0 +1,281 @@ +from __future__ import annotations + +import numpy as np +import pandas as pd + +from shared.noise_analysis import ( + DOCUMENTED_LOCAL_NOISE_ANALYSIS_POST_ROWS, + DOCUMENTED_LOCAL_NOISE_ANALYSIS_PRE_ROWS, + first_detectable_time_from_baseline, + quantify_analysis, +) +from shared.noise_snr import hash_string + + +_POSITION_COLUMN = "Position" +_VELOCITY_COLUMN = "Velocity" +_FORCE_COLUMN = "Hard_Stop_f" + + +def _scaled_noise_profile( + profile: dict[str, float], + *, + scale: float, + unscaled_keys: set[str] | None = None, +) -> dict[str, float]: + fixed = unscaled_keys or set() + return { + key: float(value) if key in fixed else float(value) * float(scale) + for key, value in profile.items() + } + + +LOW = { + "position_base_sigma_scale": 0.002, + "position_drift_sigma_scale": 0.002, + "position_event_sigma_scale": 0.002, + "position_quant_step_scale": 0.002, + "position_quant_step_floor": 1e-4, + "velocity_base_sigma_scale": 0.002, + "velocity_hetero_sigma_scale": 0.002, + "velocity_drift_sigma_scale": 0.002, + "velocity_event_sigma_scale": 0.002, + "force_base_sigma_scale": 0.002, + "force_hetero_sigma_scale": 0.002, + "force_event_sigma_scale": 0.002, +} + +HIGH_SCALE = 4.0 +HIGH = _scaled_noise_profile( + LOW, + scale=HIGH_SCALE, +) + +NOISE_DICT = {"low": LOW, "high": HIGH} +SNR_THR_DICT = { + "low": {"global": [-10000, -10000,-10000], "local": [-10000, -10000, -10000]}, + "high": {"global": [-10000, -10000,-10000], "local": [-10000, -10000, -10000]}, +} +_MAX_NOISE_RESAMPLE_ATTEMPTS = 25 + + +def _analysis_meets_thresholds( + noise_analysis: dict[str, list[float | str | None]], + *, + noise_level: str, +) -> bool: + thresholds = SNR_THR_DICT[noise_level] + for scope in ("global", "local"): + values = noise_analysis.get(scope, []) + limit_values = thresholds.get(scope, []) + for idx, raw_value in enumerate(values): + if raw_value is None or idx >= len(limit_values): + continue + value = float(raw_value) + if np.isfinite(value) and value < float(limit_values[idx]): + return False + return True + + +def _rng(seed: int, key: str) -> np.random.Generator: + derived = (int(seed) ^ hash_string(key)) & 0xFFFFFFFF + return np.random.default_rng(derived) + + +def _values(df: pd.DataFrame, column: str) -> np.ndarray: + return pd.to_numeric(df[column], errors="coerce").to_numpy(dtype=float) + + +def _finite_scale(values: np.ndarray) -> float: + finite = values[np.isfinite(values)] + if finite.size == 0: + return 1.0 + spread = float(np.nanmax(finite) - np.nanmin(finite)) + rms = float(np.sqrt(np.mean(finite**2))) + return max(spread, rms, 1e-6) + + +def _coefficients(profile: str) -> dict[str, float]: + normalized = str(profile or "low").strip().lower() + if normalized == "low": + return LOW + if normalized == "high": + return HIGH + raise ValueError(f"Unknown noise profile '{profile}'. Expected 'low' or 'high'.") + + +def _smooth(values: np.ndarray) -> np.ndarray: + kernel = np.array([0.2, 0.3, 0.3, 0.2], dtype=float) + return np.convolve(values, kernel, mode="same") + + +def _drift(rng: np.random.Generator, n: int, scale: float) -> np.ndarray: + if n <= 0 or scale <= 0.0: + return np.zeros(n, dtype=float) + return _smooth(_smooth(rng.normal(0.0, scale, size=n))) + + +def _bounce_mask(position: np.ndarray, velocity: np.ndarray) -> np.ndarray: + n = min(position.size, velocity.size) + out = np.zeros(n, dtype=bool) + if n == 0: + return out + floor = float(np.nanmin(position[np.isfinite(position)])) if np.isfinite(position).any() else 0.0 + for idx in range(1, n): + if not np.isfinite(velocity[idx - 1]) or not np.isfinite(velocity[idx]): + continue + if not np.isfinite(position[idx]): + continue + is_bounce = velocity[idx - 1] < 0.0 and velocity[idx] > 0.0 and position[idx] <= floor + 0.1 + if is_bounce: + lo = max(0, idx - 2) + hi = min(n, idx + 3) + out[lo:hi] = True + return out + + +def _add_noise_once(df: pd.DataFrame, seed: int = 0, profile: str = "low") -> pd.DataFrame: + coeffs = _coefficients(profile) + out = df.copy() + if ( + _POSITION_COLUMN not in out.columns + and _VELOCITY_COLUMN not in out.columns + and _FORCE_COLUMN not in out.columns + ): + return out + + position = ( + _values(out, _POSITION_COLUMN) + if _POSITION_COLUMN in out.columns + else np.array([], dtype=float) + ) + velocity = ( + _values(out, _VELOCITY_COLUMN) + if _VELOCITY_COLUMN in out.columns + else np.array([], dtype=float) + ) + bounces = _bounce_mask(position, velocity) + + if _POSITION_COLUMN in out.columns: + values = position + scale = _finite_scale(values) + rng = _rng(seed, _POSITION_COLUMN) + noisy = values.copy() + noisy += rng.normal( + 0.0, + coeffs["position_base_sigma_scale"] * scale, + size=values.size, + ) + noisy += _drift(rng, values.size, coeffs["position_drift_sigma_scale"] * scale) + if bounces.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["position_event_sigma_scale"] * scale, + size=values.size, + ) + * bounces.astype(float) + ) + quant_step = max( + coeffs["position_quant_step_scale"] * scale, + coeffs["position_quant_step_floor"], + ) + noisy = np.round(noisy / quant_step) * quant_step + out[_POSITION_COLUMN] = np.maximum(noisy, 0.0) + + if _VELOCITY_COLUMN in out.columns: + values = velocity + scale = _finite_scale(values) + rng = _rng(seed, _VELOCITY_COLUMN) + speed = np.abs(values) + ref = float(np.nanmedian(speed[np.isfinite(speed)])) if np.isfinite(speed).any() else 0.0 + sigma = ( + coeffs["velocity_base_sigma_scale"] * scale + + coeffs["velocity_hetero_sigma_scale"] * np.maximum(speed, ref) + ) + noisy = values.copy() + noisy += rng.normal(0.0, sigma, size=values.size) + noisy += _drift(rng, values.size, coeffs["velocity_drift_sigma_scale"] * scale) + if bounces.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["velocity_event_sigma_scale"] * scale, + size=values.size, + ) + * bounces.astype(float) + ) + out[_VELOCITY_COLUMN] = noisy + + if _FORCE_COLUMN in out.columns: + values = _values(out, _FORCE_COLUMN) + scale = _finite_scale(values) + rng = _rng(seed, _FORCE_COLUMN) + magnitude = np.abs(values) + ref = ( + float(np.nanmedian(magnitude[np.isfinite(magnitude)])) + if np.isfinite(magnitude).any() + else 0.0 + ) + sigma = ( + coeffs["force_base_sigma_scale"] * scale + + coeffs["force_hetero_sigma_scale"] * np.maximum(magnitude, ref) + ) + noisy = values.copy() + noisy += rng.normal(0.0, sigma, size=values.size) + if bounces.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["force_event_sigma_scale"] * scale, + size=values.size, + ) + * bounces.astype(float) + ) + out[_FORCE_COLUMN] = noisy + + return out + + +def quantify_noise( + clean: pd.DataFrame, + noisy: pd.DataFrame, + baseline: pd.DataFrame | None, +) -> dict[str, list[float | str | None]]: + first_diff = first_detectable_time_from_baseline(clean, baseline) + analysis = quantify_analysis( + clean, + noisy, + reference_df=baseline, + first_diff=first_diff, + local_pre_rows=DOCUMENTED_LOCAL_NOISE_ANALYSIS_PRE_ROWS, + local_post_rows=DOCUMENTED_LOCAL_NOISE_ANALYSIS_POST_ROWS, + ) + if first_diff is None or "local" not in analysis: + analysis["local"] = [None] * len(analysis.get("global", [])) + return analysis + + +def add_noise( + clean: pd.DataFrame, + baseline: pd.DataFrame | None, + seed: int = 0, + noise_level: str = "low", +) -> tuple[pd.DataFrame, dict[str, list[float | str | None]]]: + normalized = str(noise_level or "low").strip().lower() + if normalized not in NOISE_DICT: + raise ValueError(f"Unknown noise level '{noise_level}'. Expected 'low' or 'high'.") + current_seed = int(seed) + for _attempt in range(_MAX_NOISE_RESAMPLE_ATTEMPTS + 1): + noisy_df = _add_noise_once(clean, seed=current_seed, profile=normalized) + noise_analysis = quantify_noise(clean, noisy_df, baseline) + if _analysis_meets_thresholds(noise_analysis, noise_level=normalized): + return noisy_df, noise_analysis + current_seed += 1000 + raise RuntimeError( + f"Could not satisfy minimum SNR thresholds for noise level '{normalized}' " + f"after {_MAX_NOISE_RESAMPLE_ATTEMPTS + 1} attempts." + ) + + +__all__ = ["HIGH", "LOW", "NOISE_DICT", "SNR_THR_DICT", "add_noise", "quantify_noise"] diff --git a/questions/BallDrop/questions.json b/questions/BallDrop/questions.json new file mode 100644 index 0000000000000000000000000000000000000000..3f28c6b830c7407e44d63b39b076fe975acc2fab --- /dev/null +++ b/questions/BallDrop/questions.json @@ -0,0 +1,28555 @@ +{ + "version": 8, + "questions": { + "frost_01234-anchor_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "fern_01234-anchor_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "gentle_01234-anchor_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "frost_01234-cloud_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "fern_01234-cloud_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "gentle_01234-cloud_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "frost_01234-pine_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "fern_01234-pine_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "gentle_01234-pine_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "frost_01234-prairie_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "fern_01234-prairie_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "gentle_01234-prairie_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "frost_01234-spruce_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "fern_01234-spruce_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "gentle_01234-spruce_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "frost_01234-comet_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "fern_01234-comet_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "gentle_01234-comet_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "frost_01234-meadow_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "fern_01234-meadow_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "gentle_01234-meadow_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "frost_01234-river_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "fern_01234-river_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "gentle_01234-river_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "frost_01234-harbor_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "fern_01234-harbor_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "gentle_01234-harbor_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "frost_01234-willow_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "fern_01234-willow_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "gentle_01234-willow_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "frost_01234-flame_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "fern_01234-flame_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "gentle_01234-flame_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "frost_01234-orbit_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "fern_01234-orbit_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "gentle_01234-orbit_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "frost_01234-trail_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "fern_01234-trail_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "gentle_01234-trail_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "frost_01234-island_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "fern_01234-island_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "gentle_01234-island_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "frost_01234-glade_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "fern_01234-glade_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "gentle_01234-glade_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "frost_01234-canyon_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "fern_01234-canyon_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "gentle_01234-canyon_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "frost_01234-ember_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "fern_01234-ember_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "gentle_01234-ember_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "frost_01234-tide_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "fern_01234-tide_0": { + "question_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_1": { + "question_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_2": { + "question_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_3": { + "question_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_4": { + "question_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "gentle_01234-tide_0": { + "question_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_1": { + "question_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_2": { + "question_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_3": { + "question_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_4": { + "question_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "frost_01234-crest_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-star_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-forest_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a 1D vertical point-mass ball model under gravity.\nWhile the ball is above the ground, its motion is governed by gravity and quadratic air drag acting opposite the direction of travel, and the position evolves according to the current velocity.\nGround interaction is modeled with a restitution-based hard-stop law.\nWhen the ball reaches the ground with impact speed at or above 0.5 m/s, the impact is treated as instantaneous and the post-impact speed is set by the coefficient of restitution e in [0,1].\nThe rebound points upward and its magnitude equals e times the pre-impact speed.\nWhen the impact speed is below the threshold, the ball does not rebound and instead enters static contact at the ground.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-lagoon_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-quartz_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-dawn_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: ball height\ncol2: ball velocity\ncol3: contact impulse (N*s)\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution", + "mass", + "drag coefficient", + "gravity acceleration", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution\", \"mass\", \"drag coefficient\", \"gravity acceleration\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-summit_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-aurora_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-valley_0": { + "question_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_1": { + "question_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_2": { + "question_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_3": { + "question_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_4": { + "question_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + } + }, + "label_int_mapping": { + "coefficient of restitution": 0, + "mass": 1, + "drag coefficient": 2, + "gravity acceleration": 3, + "no parameter change": 4 + }, + "ground_truth_information": { + "pattern_observed": { + "bouncing_events": "List of impact times when the ball bounces on the ground.", + "apex_heights": "List of all the apex heights after each impact.", + "is_terminal_velocity": "Time intervals during which the ball is approximately at terminal velocity.", + "ball_on_the_ground": "Time from which the ball remains at rest on the ground, or JSON null if this never occurs.", + "pre_and_post_impact_velocities": "List of estimated pre-impact and post-impact velocity pairs for each detected bounce.", + "bounce_split_free_flight_parameter_estimates": "Per-bounce free-flight parameter estimates before and after a detected bounce.", + "impulse_over_diff_speed": "Ground reaction force divided by output velocity minus input velocity.", + "gravity_coefficient_estimates": "List of gravity and drag-coefficient estimates before and after free-flight splits.", + "discontinuity_in_speed_during_trajectory": "Change-point dictionaries detected during free flight where the velocity pattern changes abruptly." + }, + "interventions": { + "4d36b3bd1e524a49b337e2ac9a97206b": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.6, + "initial_height": 5.087829611649644, + "mass": 0.5, + "gravity": 5, + "initial_velocity": -6 + }, + "intervention_time": 2.55, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "00ae112c60ec4dd1acd1bacbec2560c9": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.6, + "initial_height": 5.087829611649644, + "mass": 0.5, + "gravity": 5, + "initial_velocity": -6 + }, + "intervention_time": 2.55, + "changed_parameter": "restitution", + "new_value": 0.9175361531158589, + "first_diff": [ + 2.9049999713897705, + 2.9049999713897705, + 4.255000114440918 + ] + }, + "61cd70ea73e1497f981510b030571fe5": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.6, + "initial_height": 5.087829611649644, + "mass": 0.5, + "gravity": 5, + "initial_velocity": -6 + }, + "intervention_time": 2.55, + "changed_parameter": "restitution", + "new_value": 0.1918952123886838, + "first_diff": [ + 2.9049999713897705, + 2.9049999713897705, + 4.255000114440918 + ] + }, + "439d72f79dac13bab27ea2190255b070": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.9, + "initial_height": 0.19122167512091837, + "mass": 2, + "gravity": 3, + "initial_velocity": -1.5404391977660747 + }, + "intervention_time": 1.84, + "changed_parameter": "mass", + "new_value": 9.794561166210329, + "first_diff": [ + 2.1449999809265137, + 2.132499933242798, + 2.005000114440918 + ] + }, + "adfc9499d2b940ef845993221934f87a": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.9, + "initial_height": 0.19122167512091837, + "mass": 2, + "gravity": 3, + "initial_velocity": -1.5404391977660747 + }, + "intervention_time": 1.84, + "changed_parameter": "mass", + "new_value": 0.5047181537294128, + "first_diff": [ + 1.9600000381469727, + 1.9275000095367432, + 2.007499933242798 + ] + }, + "ca218ab4f28747f6b38fc0149e36eb9b": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.9, + "initial_height": 0.19122167512091837, + "mass": 2, + "gravity": 3, + "initial_velocity": -1.5404391977660747 + }, + "intervention_time": 1.84, + "changed_parameter": "drag_coeff", + "new_value": 8.837209094117922, + "first_diff": [ + 2.0274999141693115, + 1.8550000190734863, + 2.007499933242798 + ] + }, + "c526889cd0b44bdb9efc5223a2d024a6": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.9, + "initial_height": 0.19122167512091837, + "mass": 2, + "gravity": 3, + "initial_velocity": -1.5404391977660747 + }, + "intervention_time": 1.84, + "changed_parameter": "restitution", + "new_value": 0.44470333424622166, + "first_diff": [ + 2.009999990463257, + 2.007499933242798, + 2.380000114440918 + ] + }, + "e415e80bfb684ba0a7eec7c7a02ce180": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.9, + "initial_height": 0.19122167512091837, + "mass": 2, + "gravity": 3, + "initial_velocity": -1.5404391977660747 + }, + "intervention_time": 1.84, + "changed_parameter": "gravity", + "new_value": 10.457823662148648, + "first_diff": [ + 1.8825000524520874, + 1.8450000286102295, + 1.9600000381469727 + ] + }, + "2e34f5dce75f46be81d22e0deae7b30f": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.2835059661335366, + "mass": 0.55, + "gravity": 3.6081073372402073, + "initial_velocity": -4.641207 + }, + "intervention_time": 3.4, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "4c1e9b4e56e32073502d64f3e9b72356": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.2835059661335366, + "mass": 0.55, + "gravity": 3.6081073372402073, + "initial_velocity": -4.641207 + }, + "intervention_time": 3.4, + "changed_parameter": "mass", + "new_value": 8.63353872781877, + "first_diff": [ + null, + null, + 3.672499895095825 + ] + }, + "08a23a9f66584605a0ce62ddaf3fcdd8": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.2835059661335366, + "mass": 0.55, + "gravity": 3.6081073372402073, + "initial_velocity": -4.641207 + }, + "intervention_time": 3.4, + "changed_parameter": "drag_coeff", + "new_value": 2.349529379560213, + "first_diff": [ + 3.442500114440918, + 3.4024999141693115, + 3.672499895095825 + ] + }, + "39330e4b3b284bfa859687cd7abf9b14": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.2835059661335366, + "mass": 0.55, + "gravity": 3.6081073372402073, + "initial_velocity": -4.641207 + }, + "intervention_time": 3.4, + "changed_parameter": "gravity", + "new_value": 10.813821134383046, + "first_diff": [ + 3.509999990463257, + 3.434999942779541, + 3.637500047683716 + ] + }, + "0f159f29c1004403a466c6bea03d7d02": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.99, + "initial_height": 5.000501193188524, + "mass": 0.5, + "gravity": 2.249213939930252, + "initial_velocity": 0 + }, + "intervention_time": 4.6, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "91ddb1413c4fbf36279927c5338fc650": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.99, + "initial_height": 5.000501193188524, + "mass": 0.5, + "gravity": 2.249213939930252, + "initial_velocity": 0 + }, + "intervention_time": 4.6, + "changed_parameter": "mass", + "new_value": 13.48447237544466, + "first_diff": [ + 4.892499923706055, + 4.690000057220459, + 4.882500171661377 + ] + }, + "8845ac4101424920bfad9f1eebc21c1c": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.99, + "initial_height": 5.000501193188524, + "mass": 0.5, + "gravity": 2.249213939930252, + "initial_velocity": 0 + }, + "intervention_time": 4.6, + "changed_parameter": "drag_coeff", + "new_value": 3.5160783492202023, + "first_diff": [ + 4.702499866485596, + 4.614999771118164, + 4.90500020980835 + ] + }, + "9d38efbdeee64affb143c411f6081d62": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.99, + "initial_height": 5.000501193188524, + "mass": 0.5, + "gravity": 2.249213939930252, + "initial_velocity": 0 + }, + "intervention_time": 4.6, + "changed_parameter": "restitution", + "new_value": 0.6051261498628784, + "first_diff": [ + 4.90749979019165, + 4.90500020980835, + 5.454999923706055 + ] + }, + "bc7a8ee50c0b4cd3bf478e3b3f4230dd": { + "initial_parameters": { + "drag_coeff": 0.5, + "restitution": 0.99, + "initial_height": 5.000501193188524, + "mass": 0.5, + "gravity": 2.249213939930252, + "initial_velocity": 0 + }, + "intervention_time": 4.6, + "changed_parameter": "gravity", + "new_value": 6.574669821291934, + "first_diff": [ + 4.672500133514404, + 4.607500076293945, + 4.815000057220459 + ] + }, + "cf90c926e3d14e669525f9577984bec0": { + "initial_parameters": { + "drag_coeff": 1.6927004112100619, + "restitution": 0.9777159846932046, + "initial_height": 2.559430249392824, + "mass": 19.433796184358354, + "gravity": 3.9251598494050404, + "initial_velocity": -7.178092495174859 + }, + "intervention_time": 2.2, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "a26263883fa446e49b5b921cfdf3944d": { + "initial_parameters": { + "drag_coeff": 1.6927004112100619, + "restitution": 0.9777159846932046, + "initial_height": 2.559430249392824, + "mass": 19.433796184358354, + "gravity": 3.9251598494050404, + "initial_velocity": -7.178092495174859 + }, + "intervention_time": 2.2, + "changed_parameter": "drag_coeff", + "new_value": 20.345638773834295, + "first_diff": [ + 2.507499933242798, + 2.2300000190734863, + 3.262500047683716 + ] + }, + "8f30e32b7b6b4d04915b710074abb452": { + "initial_parameters": { + "drag_coeff": 1.6927004112100619, + "restitution": 0.9777159846932046, + "initial_height": 2.559430249392824, + "mass": 19.433796184358354, + "gravity": 3.9251598494050404, + "initial_velocity": -7.178092495174859 + }, + "intervention_time": 2.2, + "changed_parameter": "restitution", + "new_value": 0.45321293193575113, + "first_diff": [ + 3.262500047683716, + 3.262500047683716, + 3.262500047683716 + ] + }, + "cfc1dde0ab3fe2803be3d63dce48d931": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 5.23660106060696, + "mass": 2.1, + "gravity": 10, + "initial_velocity": -5.395737013403933 + }, + "intervention_time": 3.75, + "changed_parameter": "mass", + "new_value": 8.516808544618655, + "first_diff": [ + 4.019999980926514, + 4.072500228881836, + 4.017499923706055 + ] + }, + "6c0a4bc64f2f4ff99f76cd41e1ab474d": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 5.23660106060696, + "mass": 2.1, + "gravity": 10, + "initial_velocity": -5.395737013403933 + }, + "intervention_time": 3.75, + "changed_parameter": "mass", + "new_value": 0.6277699747880112, + "first_diff": [ + 3.922499895095825, + 3.8524999618530273, + 4.022500038146973 + ] + }, + "631af851dff9494ca7d21f21dac03a3e": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 5.23660106060696, + "mass": 2.1, + "gravity": 10, + "initial_velocity": -5.395737013403933 + }, + "intervention_time": 3.75, + "changed_parameter": "drag_coeff", + "new_value": 3.422416078243343, + "first_diff": [ + 3.8575000762939453, + 3.7699999809265137, + 4.022500038146973 + ] + }, + "b7c15bc7416f4169b9f2a7166c4865e3": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 5.23660106060696, + "mass": 2.1, + "gravity": 10, + "initial_velocity": -5.395737013403933 + }, + "intervention_time": 3.75, + "changed_parameter": "gravity", + "new_value": 13.959638893033452, + "first_diff": [ + 3.867500066757202, + 3.7750000953674316, + 3.994999885559082 + ] + }, + "768257b526264466aff05fd9c41ab915": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 5.23660106060696, + "mass": 2.1, + "gravity": 10, + "initial_velocity": -5.395737013403933 + }, + "intervention_time": 3.75, + "changed_parameter": "gravity", + "new_value": 5.6119501078366625, + "first_diff": [ + 3.865000009536743, + 3.7699999809265137, + 4.022500038146973 + ] + }, + "f80e701fc3124cb1bf1cf9a5dfb89792": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 3.04831918243134, + "mass": 0.95, + "gravity": 8.108429979970072, + "initial_velocity": 1.146890476513768 + }, + "intervention_time": 2.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "001063c5dfa248dbbba122ff6743af5f": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 3.04831918243134, + "mass": 0.95, + "gravity": 8.108429979970072, + "initial_velocity": 1.146890476513768 + }, + "intervention_time": 2.0, + "changed_parameter": "drag_coeff", + "new_value": 2.4600786759019475, + "first_diff": [ + 2.2149999141693115, + 2.015000104904175, + 2.692500114440918 + ] + }, + "a6e56514770442afb6af22c381df3a0a": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 3.04831918243134, + "mass": 0.95, + "gravity": 8.108429979970072, + "initial_velocity": 1.146890476513768 + }, + "intervention_time": 2.0, + "changed_parameter": "restitution", + "new_value": 0.38495003656379767, + "first_diff": [ + 2.692500114440918, + 2.692500114440918, + 2.692500114440918 + ] + }, + "efeae73360874aefb8d1d01031db168a": { + "initial_parameters": { + "drag_coeff": 1.1, + "restitution": 0.975, + "initial_height": 2.5482589870392878, + "mass": 1.1, + "gravity": 2, + "initial_velocity": -8.783110990262688 + }, + "intervention_time": 3.05, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "a1c583110649c9166d2b8cc4ab82eca6": { + "initial_parameters": { + "drag_coeff": 1.1, + "restitution": 0.975, + "initial_height": 2.5482589870392878, + "mass": 1.1, + "gravity": 2, + "initial_velocity": -8.783110990262688 + }, + "intervention_time": 3.05, + "changed_parameter": "mass", + "new_value": 5.313865987665894, + "first_diff": [ + 3.2274999618530273, + 3.130000114440918, + 3.0999999046325684 + ] + }, + "3a5ad8a0c7a84fd3a5baaec0eee7a126": { + "initial_parameters": { + "drag_coeff": 1.1, + "restitution": 0.975, + "initial_height": 2.5482589870392878, + "mass": 1.1, + "gravity": 2, + "initial_velocity": -8.783110990262688 + }, + "intervention_time": 1.05, + "changed_parameter": "drag_coeff", + "new_value": 29.42960583053023, + "first_diff": [ + 1.0824999809265137, + 1.0525000095367432, + 2.184999942779541 + ] + }, + "ab2618c21bea42a5b9d58b93173c082c": { + "initial_parameters": { + "drag_coeff": 1.1, + "restitution": 0.975, + "initial_height": 2.5482589870392878, + "mass": 1.1, + "gravity": 2, + "initial_velocity": -8.783110990262688 + }, + "intervention_time": 3.05, + "changed_parameter": "restitution", + "new_value": 0.1617564171022836, + "first_diff": [ + 3.0999999046325684, + 3.0999999046325684, + 3.0999999046325684 + ] + }, + "b5641f9340fa4cd08749657345a1bedf": { + "initial_parameters": { + "drag_coeff": 1.1, + "restitution": 0.975, + "initial_height": 2.5482589870392878, + "mass": 1.1, + "gravity": 2, + "initial_velocity": -8.783110990262688 + }, + "intervention_time": 3.05, + "changed_parameter": "gravity", + "new_value": 5.362968547089867, + "first_diff": [ + 3.2200000286102295, + 3.1449999809265137, + 3.0950000286102295 + ] + }, + "7d1b3f1a85c049ff9b8bd7c7a18d4041": { + "initial_parameters": { + "drag_coeff": 6, + "restitution": 0.98, + "initial_height": 0.377251815045354, + "mass": 3.2, + "gravity": 6, + "initial_velocity": -8.558628604874482 + }, + "intervention_time": 0.4, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "dcbbcb9aaba84f95b6e6c410a573039a": { + "initial_parameters": { + "drag_coeff": 6, + "restitution": 0.98, + "initial_height": 0.377251815045354, + "mass": 3.2, + "gravity": 6, + "initial_velocity": -8.558628604874482 + }, + "intervention_time": 0.12, + "changed_parameter": "drag_coeff", + "new_value": 18.435342213785837, + "first_diff": [ + 0.1599999964237213, + 0.125, + 0.8224999904632568 + ] + }, + "12468fab1955463b813e11f74847029b": { + "initial_parameters": { + "drag_coeff": 6, + "restitution": 0.98, + "initial_height": 0.377251815045354, + "mass": 3.2, + "gravity": 6, + "initial_velocity": -8.558628604874482 + }, + "intervention_time": 1.0, + "changed_parameter": "restitution", + "new_value": 0.5770846171020974, + "first_diff": [ + 1.3825000524520874, + 1.3799999952316284, + 1.6100000143051147 + ] + }, + "3df1d923d25b44d6826383b65210f7eb": { + "initial_parameters": { + "drag_coeff": 6, + "restitution": 0.98, + "initial_height": 0.377251815045354, + "mass": 3.2, + "gravity": 6, + "initial_velocity": -8.558628604874482 + }, + "intervention_time": 2.05, + "changed_parameter": "gravity", + "new_value": 2.545982032538134, + "first_diff": [ + 2.127500057220459, + 2.072499990463257, + 2.055000066757202 + ] + }, + "42e0669504784adea7e79a929f5c5495": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.955, + "initial_height": 1.1423535246605752, + "mass": 0.9, + "gravity": 2.410480825425651, + "initial_velocity": 1.1238720870970216 + }, + "intervention_time": 2.5, + "changed_parameter": "mass", + "new_value": 11.776977066513098, + "first_diff": [ + 2.944999933242798, + 2.8499999046325684, + 3.109999895095825 + ] + }, + "831770be078648aebe505cfc3add6f65": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.955, + "initial_height": 1.1423535246605752, + "mass": 0.9, + "gravity": 2.410480825425651, + "initial_velocity": 1.1238720870970216 + }, + "intervention_time": 2.5, + "changed_parameter": "drag_coeff", + "new_value": 14.397491171977954, + "first_diff": [ + 2.6624999046325684, + 2.507499933242798, + 3.134999990463257 + ] + }, + "9469a80f76084842933ee512968eadc2": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.955, + "initial_height": 1.1423535246605752, + "mass": 0.9, + "gravity": 2.410480825425651, + "initial_velocity": 1.1238720870970216 + }, + "intervention_time": 2.5, + "changed_parameter": "restitution", + "new_value": 0.5799368112671375, + "first_diff": [ + 3.134999990463257, + 3.134999990463257, + 3.932499885559082 + ] + }, + "2711a07092314ea681315ca82e5bc51b": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.955, + "initial_height": 1.1423535246605752, + "mass": 0.9, + "gravity": 2.410480825425651, + "initial_velocity": 1.1238720870970216 + }, + "intervention_time": 2.5, + "changed_parameter": "gravity", + "new_value": 7.667976719939459, + "first_diff": [ + 2.609999895095825, + 2.502500057220459, + 2.887500047683716 + ] + }, + "81cb67a56e3b53dabdc75ab3749d1a65": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 1.5672514750709976, + "mass": 2.1, + "gravity": 9.030781692753983, + "initial_velocity": 0.5883090079098228 + }, + "intervention_time": 3.8625, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "39766784659f583399ef48bbd86c3a85": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 1.5672514750709976, + "mass": 2.1, + "gravity": 9.030781692753983, + "initial_velocity": 0.5883090079098228 + }, + "intervention_time": 3.8625, + "changed_parameter": "drag_coeff", + "new_value": 4.68, + "first_diff": [ + 3.890000104904175, + 3.869999885559082, + 4.327499866485596 + ] + }, + "6c7ac291a99852eea1bd5fb4a619a097": { + "initial_parameters": { + "drag_coeff": 0.25, + "restitution": 0.95, + "initial_height": 1.5672514750709976, + "mass": 2.1, + "gravity": 9.030781692753983, + "initial_velocity": 0.5883090079098228 + }, + "intervention_time": 3.8625, + "changed_parameter": "gravity", + "new_value": 4.6382141132264, + "first_diff": [ + 3.930000066757202, + 3.890000104904175, + 4.452499866485596 + ] + }, + "9583b6b3a5c548c0893c647f3350dd5a": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 4.735490727300136, + "mass": 1.7, + "gravity": 4, + "initial_velocity": -4.35 + }, + "intervention_time": 3.55, + "changed_parameter": "mass", + "new_value": 10.651685301728385, + "first_diff": [ + null, + null, + 4.382500171661377 + ] + }, + "a58f39410c8a58289fe81de42778819a": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 4.735490727300136, + "mass": 1.7, + "gravity": 4, + "initial_velocity": -4.35 + }, + "intervention_time": 3.55, + "changed_parameter": "mass", + "new_value": 0.5368939355664853, + "first_diff": [ + null, + null, + 4.382500171661377 + ] + }, + "cf5492d1ceb754e6af4c83df1d1e646f": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 4.735490727300136, + "mass": 1.7, + "gravity": 4, + "initial_velocity": -4.35 + }, + "intervention_time": 3.55, + "changed_parameter": "drag_coeff", + "new_value": 2.6087754596616497, + "first_diff": [ + 3.7100000381469727, + 3.557499885559082, + 4.382500171661377 + ] + }, + "568b3e9ed49658d7abaae622daac8cef": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 4.735490727300136, + "mass": 1.7, + "gravity": 4, + "initial_velocity": -4.35 + }, + "intervention_time": 3.55, + "changed_parameter": "restitution", + "new_value": 0.31169602915319944, + "first_diff": [ + 4.382500171661377, + 4.382500171661377, + 4.382500171661377 + ] + }, + "2e86956a56055b2aa94c0030af2f2f74": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 4.735490727300136, + "mass": 1.7, + "gravity": 4, + "initial_velocity": -4.35 + }, + "intervention_time": 3.55, + "changed_parameter": "gravity", + "new_value": 11.788754110132022, + "first_diff": [ + 3.762500047683716, + 3.575000047683716, + 4.162499904632568 + ] + }, + "4757e7217e7a56d693494ebf2f740d43": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.925, + "initial_height": 2.553571, + "mass": 3.2, + "gravity": 6.687749549282168, + "initial_velocity": -7.278900705603197 + }, + "intervention_time": 5.002025, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "c3eb6b6c9abe403faec7dd1a28b4378e": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.925, + "initial_height": 2.553571, + "mass": 3.2, + "gravity": 6.687749549282168, + "initial_velocity": -7.278900705603197 + }, + "intervention_time": 5.002025, + "changed_parameter": "mass", + "new_value": 16.164044023277967, + "first_diff": [ + null, + null, + 5.289999961853027 + ] + }, + "e799ab902f045c1d9c29b64ced889454": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.925, + "initial_height": 2.553571, + "mass": 3.2, + "gravity": 6.687749549282168, + "initial_velocity": -7.278900705603197 + }, + "intervention_time": 5.002025, + "changed_parameter": "mass", + "new_value": 1.1926453071916132, + "first_diff": [ + null, + null, + 5.289999961853027 + ] + }, + "ccf1034ad54459debd65aff402728604": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.925, + "initial_height": 2.553571, + "mass": 3.2, + "gravity": 6.687749549282168, + "initial_velocity": -7.278900705603197 + }, + "intervention_time": 5.002025, + "changed_parameter": "drag_coeff", + "new_value": 2.8853311551913903, + "first_diff": [ + 5.077499866485596, + 5.010000228881836, + 5.289999961853027 + ] + }, + "a236c7b60f6d489cad9fdc8ef847a425": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.925, + "initial_height": 2.553571, + "mass": 3.2, + "gravity": 6.687749549282168, + "initial_velocity": -7.278900705603197 + }, + "intervention_time": 5.002025, + "changed_parameter": "restitution", + "new_value": 0.5450454856839533, + "first_diff": [ + 5.289999961853027, + 5.289999961853027, + 6.602499961853027 + ] + }, + "5f875cc969b55f368b95aee1a44d01e5": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.925, + "initial_height": 2.553571, + "mass": 3.2, + "gravity": 6.687749549282168, + "initial_velocity": -7.278900705603197 + }, + "intervention_time": 5.002025, + "changed_parameter": "gravity", + "new_value": 2.6069533260095623, + "first_diff": [ + 5.159999847412109, + 5.079999923706055, + 5.289999961853027 + ] + }, + "4373eb3bac935c4e8b80d82e28f50ffe": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.7338753569687695, + "mass": 3.2, + "gravity": 10, + "initial_velocity": 0.6386524240074678 + }, + "intervention_time": 8.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "a7a514d5d4565e85b4f48ad78735021e": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.7338753569687695, + "mass": 3.2, + "gravity": 10, + "initial_velocity": 0.6386524240074678 + }, + "intervention_time": 8.0, + "changed_parameter": "mass", + "new_value": 1.3136542864355079, + "first_diff": [ + null, + null, + 8.357500076293945 + ] + }, + "1b4abf9677a956e4bb90051b4494f4b3": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.7338753569687695, + "mass": 3.2, + "gravity": 10, + "initial_velocity": 0.6386524240074678 + }, + "intervention_time": 8.0, + "changed_parameter": "drag_coeff", + "new_value": 2.8933173053054624, + "first_diff": [ + 8.145000457763672, + 8.02750015258789, + 8.357500076293945 + ] + }, + "3e706177423c5f289117291cc571bc8d": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.7338753569687695, + "mass": 3.2, + "gravity": 10, + "initial_velocity": 0.6386524240074678 + }, + "intervention_time": 8.0, + "changed_parameter": "restitution", + "new_value": 0.38754097440304536, + "first_diff": [ + 8.357500076293945, + 8.357500076293945, + 8.357500076293945 + ] + }, + "81867907eb7c5370b8f73d2718d8398d": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.7338753569687695, + "mass": 3.2, + "gravity": 10, + "initial_velocity": 0.6386524240074678 + }, + "intervention_time": 8.0, + "changed_parameter": "gravity", + "new_value": 14.629292154771113, + "first_diff": [ + 8.142499923706055, + 8.024999618530273, + 8.3149995803833 + ] + }, + "cd023a4ad4de4316bba0e94b5bf35d8c": { + "initial_parameters": { + "drag_coeff": 0, + "restitution": 0.95, + "initial_height": 2.7338753569687695, + "mass": 3.2, + "gravity": 10, + "initial_velocity": 0.6386524240074678 + }, + "intervention_time": 8.0, + "changed_parameter": "gravity", + "new_value": 4.422625097948293, + "first_diff": [ + 8.135000228881836, + 8.017499923706055, + 8.357500076293945 + ] + }, + "0d96332e3fae53ff8c19e0445a16f304": { + "initial_parameters": { + "drag_coeff": 1.4, + "restitution": 0.95, + "initial_height": 2.353323638382498, + "mass": 3.2, + "gravity": 4.440884224888464, + "initial_velocity": -4.05 + }, + "intervention_time": 2.95, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "303d59b812a942eeac4a05f71df85ff1": { + "initial_parameters": { + "drag_coeff": 1.4, + "restitution": 0.95, + "initial_height": 2.353323638382498, + "mass": 3.2, + "gravity": 4.440884224888464, + "initial_velocity": -4.05 + }, + "intervention_time": 2.95, + "changed_parameter": "mass", + "new_value": 18.75036774171687, + "first_diff": [ + 3.444999933242798, + 3.4725000858306885, + 3.442500114440918 + ] + }, + "50b90a6df0e25ac480a3af80bb1b21bd": { + "initial_parameters": { + "drag_coeff": 1.4, + "restitution": 0.95, + "initial_height": 2.353323638382498, + "mass": 3.2, + "gravity": 4.440884224888464, + "initial_velocity": -4.05 + }, + "intervention_time": 2.95, + "changed_parameter": "mass", + "new_value": 1.4004979689452135, + "first_diff": [ + 3.3975000381469727, + 3.4600000381469727, + 3.447499990463257 + ] + }, + "00f96c534483544fbf5cb3d14e38272f": { + "initial_parameters": { + "drag_coeff": 1.4, + "restitution": 0.95, + "initial_height": 2.353323638382498, + "mass": 3.2, + "gravity": 4.440884224888464, + "initial_velocity": -4.05 + }, + "intervention_time": 2.95, + "changed_parameter": "drag_coeff", + "new_value": 25.917900982271657, + "first_diff": [ + 3.3299999237060547, + 2.9600000381469727, + 3.447499990463257 + ] + }, + "9efdb21c07b5589b9b02b808cfd7d78f": { + "initial_parameters": { + "drag_coeff": 1.4, + "restitution": 0.95, + "initial_height": 2.353323638382498, + "mass": 3.2, + "gravity": 4.440884224888464, + "initial_velocity": -4.05 + }, + "intervention_time": 2.95, + "changed_parameter": "restitution", + "new_value": 0.5646430836628626, + "first_diff": [ + 3.450000047683716, + 3.447499990463257, + 3.827500104904175 + ] + }, + "fca8a26ed97e46f5a04e19c22db27fa5": { + "initial_parameters": { + "drag_coeff": 1.5, + "restitution": 0.93, + "initial_height": 2, + "mass": 3, + "gravity": 8.662092007776046, + "initial_velocity": -4.899962321070323 + }, + "intervention_time": 1.2, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "d5e208a1a02e48d4ba8d47f9cf10a514": { + "initial_parameters": { + "drag_coeff": 1.5, + "restitution": 0.93, + "initial_height": 2, + "mass": 3, + "gravity": 8.662092007776046, + "initial_velocity": -4.899962321070323 + }, + "intervention_time": 1.2, + "changed_parameter": "restitution", + "new_value": 0.19973314825841154, + "first_diff": [ + 1.225000023841858, + 1.225000023841858, + 1.225000023841858 + ] + }, + "3e28f2d025fd4f229ce62a07f1ed72eb": { + "initial_parameters": { + "drag_coeff": 1.5, + "restitution": 0.93, + "initial_height": 2, + "mass": 3, + "gravity": 8.662092007776046, + "initial_velocity": -4.899962321070323 + }, + "intervention_time": 1.2, + "changed_parameter": "gravity", + "new_value": 4.476588329070927, + "first_diff": [ + 1.3650000095367432, + 1.2725000381469727, + 1.225000023841858 + ] + }, + "b587da7834684a6a9513f7e7b90e8a9e": { + "initial_parameters": { + "drag_coeff": 0.75, + "restitution": 0.65, + "initial_height": 4.3528672111603, + "mass": 10, + "gravity": 5, + "initial_velocity": -4.886053192563296 + }, + "intervention_time": 2.5, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "17a033a89614421b8bb520c3bb286b60": { + "initial_parameters": { + "drag_coeff": 0.75, + "restitution": 0.65, + "initial_height": 4.3528672111603, + "mass": 10, + "gravity": 5, + "initial_velocity": -4.886053192563296 + }, + "intervention_time": 2.5, + "changed_parameter": "restitution", + "new_value": 0.9616989512593546, + "first_diff": [ + 3.3524999618530273, + 3.3524999618530273, + 3.9749999046325684 + ] + }, + "919c229adaf44cd68af7629ae8feb882": { + "initial_parameters": { + "drag_coeff": 0.75, + "restitution": 0.65, + "initial_height": 4.3528672111603, + "mass": 10, + "gravity": 5, + "initial_velocity": -4.886053192563296 + }, + "intervention_time": 2.5, + "changed_parameter": "restitution", + "new_value": 0.18223726667352136, + "first_diff": [ + 3.3524999618530273, + 3.3524999618530273, + 3.3524999618530273 + ] + }, + "2348ceba66234f5f91acc812e273a3d5": { + "initial_parameters": { + "drag_coeff": 0.75, + "restitution": 0.65, + "initial_height": 4.3528672111603, + "mass": 10, + "gravity": 5, + "initial_velocity": -4.886053192563296 + }, + "intervention_time": 2.5, + "changed_parameter": "gravity", + "new_value": 11.084046659036185, + "first_diff": [ + 2.5824999809265137, + 2.512500047683716, + 2.940000057220459 + ] + }, + "7df2e9376cf654358d488e51f581b366": { + "initial_parameters": { + "drag_coeff": 0.64678, + "gravity": 8.046705, + "initial_height": 5.768982, + "initial_velocity": -3.2021722334673735, + "mass": 4.214478, + "restitution": 0.8 + }, + "intervention_time": 2.3, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "523c9218f52c57068887247f1c7a044a": { + "initial_parameters": { + "drag_coeff": 0.64678, + "gravity": 8.046705, + "initial_height": 5.768982, + "initial_velocity": -3.2021722334673735, + "mass": 4.214478, + "restitution": 0.8 + }, + "intervention_time": 2.3, + "changed_parameter": "mass", + "new_value": 0.9716096852803022, + "first_diff": [ + 2.4149999618530273, + 2.3324999809265137, + 3.015000104904175 + ] + }, + "6514e4c5d16f5fac9810e4b4e0db5d24": { + "initial_parameters": { + "drag_coeff": 0.64678, + "gravity": 8.046705, + "initial_height": 5.768982, + "initial_velocity": -3.2021722334673735, + "mass": 4.214478, + "restitution": 0.8 + }, + "intervention_time": 2.3, + "changed_parameter": "drag_coeff", + "new_value": 5.876236347369603, + "first_diff": [ + 2.3575000762939453, + 2.3125, + 2.9600000381469727 + ] + }, + "7e0f85ac769c5ab78201f4cdac386d36": { + "initial_parameters": { + "drag_coeff": 0.64678, + "gravity": 8.046705, + "initial_height": 5.768982, + "initial_velocity": -3.2021722334673735, + "mass": 4.214478, + "restitution": 0.8 + }, + "intervention_time": 2.3, + "changed_parameter": "restitution", + "new_value": 0.22349411426034568, + "first_diff": [ + 3.077500104904175, + 3.077500104904175, + 3.077500104904175 + ] + }, + "33508fdf1d3d5b62a0f61a2ab5200441": { + "initial_parameters": { + "drag_coeff": 0.64678, + "gravity": 8.046705, + "initial_height": 5.768982, + "initial_velocity": -3.2021722334673735, + "mass": 4.214478, + "restitution": 0.8 + }, + "intervention_time": 2.3, + "changed_parameter": "gravity", + "new_value": 11.542395630053782, + "first_diff": [ + 2.4124999046325684, + 2.3375000953674316, + 2.859999895095825 + ] + }, + "bb0abe5b86f652bf819eb1c65190c4c7": { + "initial_parameters": { + "drag_coeff": 1.669381, + "gravity": 5.968411, + "initial_height": 3.929558536282836, + "initial_velocity": -0.9217929832622147, + "mass": 12.723372, + "restitution": 0.377002 + }, + "intervention_time": 1.14, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "f4c8f20b5816534db655784eb8d032f6": { + "initial_parameters": { + "drag_coeff": 1.669381, + "gravity": 5.968411, + "initial_height": 3.929558536282836, + "initial_velocity": -0.9217929832622147, + "mass": 12.723372, + "restitution": 0.377002 + }, + "intervention_time": 1.14, + "changed_parameter": "mass", + "new_value": 5.923251367352265, + "first_diff": [ + 1.5549999475479126, + 1.2999999523162842, + 1.7699999809265137 + ] + }, + "c8baff863e805d089948e883182b8d93": { + "initial_parameters": { + "drag_coeff": 1.669381, + "gravity": 5.968411, + "initial_height": 3.929558536282836, + "initial_velocity": -0.9217929832622147, + "mass": 12.723372, + "restitution": 0.377002 + }, + "intervention_time": 1.14, + "changed_parameter": "gravity", + "new_value": 11.942121931692672, + "first_diff": [ + 1.1924999952316284, + 1.152500033378601, + 1.4774999618530273 + ] + }, + "63b4f438163057f9b67ff2a42d6427c6": { + "initial_parameters": { + "drag_coeff": 1.701999, + "gravity": 5.252143, + "initial_height": 4.986694001718108, + "initial_velocity": -7.207437845624812, + "mass": 11.120026, + "restitution": 0.631213 + }, + "intervention_time": 1.733008, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "8f251b7ab3515230926f69d9fcd57dd4": { + "initial_parameters": { + "drag_coeff": 1.701999, + "gravity": 5.252143, + "initial_height": 4.986694001718108, + "initial_velocity": -7.207437845624812, + "mass": 11.120026, + "restitution": 0.631213 + }, + "intervention_time": 1.733008, + "changed_parameter": "mass", + "new_value": 1.5107843654887714, + "first_diff": [ + 1.8875000476837158, + 1.7675000429153442, + 2.117500066757202 + ] + }, + "56329b0d596f5b4daa3db101e5a634e6": { + "initial_parameters": { + "drag_coeff": 1.701999, + "gravity": 5.252143, + "initial_height": 4.986694001718108, + "initial_velocity": -7.207437845624812, + "mass": 11.120026, + "restitution": 0.631213 + }, + "intervention_time": 1.733008, + "changed_parameter": "drag_coeff", + "new_value": 10.01740838979297, + "first_diff": [ + 1.902500033378601, + 1.777500033378601, + 2.117500066757202 + ] + }, + "f73e9eeb29be5e07b8c13705ff2ab2a6": { + "initial_parameters": { + "drag_coeff": 1.701999, + "gravity": 5.252143, + "initial_height": 4.986694001718108, + "initial_velocity": -7.207437845624812, + "mass": 11.120026, + "restitution": 0.631213 + }, + "intervention_time": 1.733008, + "changed_parameter": "restitution", + "new_value": 0.31307922362303187, + "first_diff": [ + 2.117500066757202, + 2.117500066757202, + 2.509999990463257 + ] + }, + "d72e4ead5b465703b756c36d29213e3f": { + "initial_parameters": { + "drag_coeff": 1.701999, + "gravity": 5.252143, + "initial_height": 4.986694001718108, + "initial_velocity": -7.207437845624812, + "mass": 11.120026, + "restitution": 0.631213 + }, + "intervention_time": 1.733008, + "changed_parameter": "restitution", + "new_value": 0.9694768737646897, + "first_diff": [ + 2.117500066757202, + 2.117500066757202, + 2.882499933242798 + ] + }, + "873c2abc53235d5eb408430e7313c0b7": { + "initial_parameters": { + "drag_coeff": 1.701999, + "gravity": 5.252143, + "initial_height": 4.986694001718108, + "initial_velocity": -7.207437845624812, + "mass": 11.120026, + "restitution": 0.631213 + }, + "intervention_time": 1.733008, + "changed_parameter": "gravity", + "new_value": 2.195440743184403, + "first_diff": [ + 1.8849999904632568, + 1.7625000476837158, + 2.117500066757202 + ] + }, + "c5333739516d5fbfb8aa22aaf11a7cc9": { + "initial_parameters": { + "drag_coeff": 1.701999, + "gravity": 5.252143, + "initial_height": 4.986694001718108, + "initial_velocity": -7.207437845624812, + "mass": 11.120026, + "restitution": 0.631213 + }, + "intervention_time": 1.733008, + "changed_parameter": "gravity", + "new_value": 11.827157205017741, + "first_diff": [ + 1.840000033378601, + 1.747499942779541, + 2.0299999713897705 + ] + }, + "34574575af47569e8859d01ec8078234": { + "initial_parameters": { + "drag_coeff": 0.001, + "gravity": 7.104208, + "initial_height": 5.709393767789777, + "initial_velocity": -9.274088419143983, + "mass": 0.962553, + "restitution": 0.4 + }, + "intervention_time": 5.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "d5e619d9d7fb52898940b8140f653c12": { + "initial_parameters": { + "drag_coeff": 0.001, + "gravity": 7.104208, + "initial_height": 5.709393767789777, + "initial_velocity": -9.274088419143983, + "mass": 0.962553, + "restitution": 0.4 + }, + "intervention_time": 1.0, + "changed_parameter": "drag_coeff", + "new_value": 4.566177580513418, + "first_diff": [ + 1.2000000476837158, + 1.0049999952316284, + 1.9674999713897705 + ] + }, + "9daf764b6051540e98b175ec43c551c4": { + "initial_parameters": { + "drag_coeff": 0.001, + "gravity": 7.104208, + "initial_height": 5.709393767789777, + "initial_velocity": -9.274088419143983, + "mass": 0.962553, + "restitution": 0.4 + }, + "intervention_time": 1.0, + "changed_parameter": "restitution", + "new_value": 0.9287932180423607, + "first_diff": [ + 1.9674999713897705, + 1.9674999713897705, + 1.9674999713897705 + ] + }, + "ca1727e8ff895ff497d01835818489e8": { + "initial_parameters": { + "drag_coeff": 0.001, + "gravity": 7.104208, + "initial_height": 5.709393767789777, + "initial_velocity": -9.274088419143983, + "mass": 0.962553, + "restitution": 0.4 + }, + "intervention_time": 1.0, + "changed_parameter": "gravity", + "new_value": 2.10069769179391, + "first_diff": [ + 1.1950000524520874, + 1.0149999856948853, + 1.9674999713897705 + ] + }, + "55cd4372b9205e238624bf60157b1250": { + "initial_parameters": { + "drag_coeff": 0.001, + "gravity": 7.104208, + "initial_height": 5.709393767789777, + "initial_velocity": -9.274088419143983, + "mass": 0.962553, + "restitution": 0.4 + }, + "intervention_time": 1.0, + "changed_parameter": "gravity", + "new_value": 11.511192308659968, + "first_diff": [ + 1.2024999856948853, + 1.0175000429153442, + 1.7074999809265137 + ] + }, + "fbe454d4b3625caaaa07aaed18d16268": { + "initial_parameters": { + "drag_coeff": 1.57644, + "gravity": 4.119865807185478, + "initial_height": 4.042281, + "initial_velocity": -6.355112879259111, + "mass": 7.127182, + "restitution": 0.6 + }, + "intervention_time": 1.0, + "changed_parameter": "mass", + "new_value": 1.808228519388863, + "first_diff": [ + 1.340000033378601, + 1.0499999523162842, + 2.0225000381469727 + ] + }, + "9c445001c2f75ddb9bd1d47993a664f2": { + "initial_parameters": { + "drag_coeff": 1.57644, + "gravity": 4.119865807185478, + "initial_height": 4.042281, + "initial_velocity": -6.355112879259111, + "mass": 7.127182, + "restitution": 0.6 + }, + "intervention_time": 1.0, + "changed_parameter": "mass", + "new_value": 14.20112857020403, + "first_diff": [ + 1.909999966621399, + 1.2324999570846558, + 2.0225000381469727 + ] + }, + "74558dec52105edfb945508fea345732": { + "initial_parameters": { + "drag_coeff": 1.57644, + "gravity": 4.119865807185478, + "initial_height": 4.042281, + "initial_velocity": -6.355112879259111, + "mass": 7.127182, + "restitution": 0.6 + }, + "intervention_time": 1.0, + "changed_parameter": "drag_coeff", + "new_value": 23.96492521655622, + "first_diff": [ + 1.122499942779541, + 1.0099999904632568, + 2.0225000381469727 + ] + }, + "93651e54ac445eba83c9f71cd64421a9": { + "initial_parameters": { + "drag_coeff": 1.57644, + "gravity": 4.119865807185478, + "initial_height": 4.042281, + "initial_velocity": -6.355112879259111, + "mass": 7.127182, + "restitution": 0.6 + }, + "intervention_time": 1.0, + "changed_parameter": "restitution", + "new_value": 0.17204142961984964, + "first_diff": [ + 2.0225000381469727, + 2.0225000381469727, + 2.0225000381469727 + ] + }, + "36aee79d125e52d18d0608ce1873c6b4": { + "initial_parameters": { + "drag_coeff": 1.57644, + "gravity": 4.119865807185478, + "initial_height": 4.042281, + "initial_velocity": -6.355112879259111, + "mass": 7.127182, + "restitution": 0.6 + }, + "intervention_time": 1.0, + "changed_parameter": "restitution", + "new_value": 0.9140617059457755, + "first_diff": [ + 2.0225000381469727, + 2.0225000381469727, + 2.7049999237060547 + ] + }, + "1dcbb406a6365e2f86382f83d5a8a508": { + "initial_parameters": { + "drag_coeff": 1.57644, + "gravity": 4.119865807185478, + "initial_height": 4.042281, + "initial_velocity": -6.355112879259111, + "mass": 7.127182, + "restitution": 0.6 + }, + "intervention_time": 1.0, + "changed_parameter": "gravity", + "new_value": 8.55880529426549, + "first_diff": [ + 1.125, + 1.0149999856948853, + 1.590000033378601 + ] + }, + "47f84b9f0967540a98380bd737333252": { + "initial_parameters": { + "drag_coeff": 0.767476, + "gravity": 5.632678, + "initial_height": 1.370150532352688, + "initial_velocity": 0.9539833224861916, + "mass": 12.265895, + "restitution": 0.9 + }, + "intervention_time": 4.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "fbee8e4e80435cb08b5438b33fb62e36": { + "initial_parameters": { + "drag_coeff": 0.767476, + "gravity": 5.632678, + "initial_height": 1.370150532352688, + "initial_velocity": 0.9539833224861916, + "mass": 12.265895, + "restitution": 0.9 + }, + "intervention_time": 4.0, + "changed_parameter": "mass", + "new_value": 3.4241500299115915, + "first_diff": [ + 4.284999847412109, + 4.130000114440918, + 4.017499923706055 + ] + }, + "6228a7c919375242b6065fac22112b1e": { + "initial_parameters": { + "drag_coeff": 0.767476, + "gravity": 5.632678, + "initial_height": 1.370150532352688, + "initial_velocity": 0.9539833224861916, + "mass": 12.265895, + "restitution": 0.9 + }, + "intervention_time": 4.0, + "changed_parameter": "restitution", + "new_value": 0.5515131064886835, + "first_diff": [ + 4.017499923706055, + 4.017499923706055, + 4.502500057220459 + ] + }, + "06c7aeb4badb5ca085c44da26f39cfa5": { + "initial_parameters": { + "drag_coeff": 0.767476, + "gravity": 5.632678, + "initial_height": 1.370150532352688, + "initial_velocity": 0.9539833224861916, + "mass": 12.265895, + "restitution": 0.9 + }, + "intervention_time": 4.0, + "changed_parameter": "gravity", + "new_value": 9.41778337564579, + "first_diff": [ + 4.110000133514404, + 4.0625, + 4.502500057220459 + ] + }, + "ba2e3cbfe88c52748c6d2c81c7a04600": { + "initial_parameters": { + "drag_coeff": 0, + "gravity": 4.85548935766852, + "initial_height": 5.307198225470252, + "initial_velocity": -9.727621534137041, + "mass": 12.268117, + "restitution": 0.5 + }, + "intervention_time": 2.3, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "77968ae18bca566cbcffba3830a3fece": { + "initial_parameters": { + "drag_coeff": 0, + "gravity": 4.85548935766852, + "initial_height": 5.307198225470252, + "initial_velocity": -9.727621534137041, + "mass": 12.268117, + "restitution": 0.5 + }, + "intervention_time": 2.3, + "changed_parameter": "mass", + "new_value": 7.344890528534786, + "first_diff": [ + null, + null, + 2.9825000762939453 + ] + }, + "18e695472b2d52058cf6deee57c49cf8": { + "initial_parameters": { + "drag_coeff": 0, + "gravity": 4.85548935766852, + "initial_height": 5.307198225470252, + "initial_velocity": -9.727621534137041, + "mass": 12.268117, + "restitution": 0.5 + }, + "intervention_time": 2.3, + "changed_parameter": "mass", + "new_value": 18.882109838865407, + "first_diff": [ + null, + null, + 2.9825000762939453 + ] + }, + "e195837b1041539c86a3733e4391198b": { + "initial_parameters": { + "drag_coeff": 0, + "gravity": 4.85548935766852, + "initial_height": 5.307198225470252, + "initial_velocity": -9.727621534137041, + "mass": 12.268117, + "restitution": 0.5 + }, + "intervention_time": 2.3, + "changed_parameter": "drag_coeff", + "new_value": 4.670273313792303, + "first_diff": [ + 2.559999942779541, + 2.3475000858306885, + 2.9825000762939453 + ] + }, + "e57e705e527d514db713ff28280d11fd": { + "initial_parameters": { + "drag_coeff": 0, + "gravity": 4.85548935766852, + "initial_height": 5.307198225470252, + "initial_velocity": -9.727621534137041, + "mass": 12.268117, + "restitution": 0.5 + }, + "intervention_time": 2.3, + "changed_parameter": "restitution", + "new_value": 0.8110485312778151, + "first_diff": [ + 2.9825000762939453, + 2.9825000762939453, + 4.235000133514404 + ] + }, + "1f176712882c5ce69f5f87c4473b6694": { + "initial_parameters": { + "drag_coeff": 0, + "gravity": 4.85548935766852, + "initial_height": 5.307198225470252, + "initial_velocity": -9.727621534137041, + "mass": 12.268117, + "restitution": 0.5 + }, + "intervention_time": 2.3, + "changed_parameter": "gravity", + "new_value": 11.67671654685926, + "first_diff": [ + 2.484999895095825, + 2.319999933242798, + 2.817500114440918 + ] + } + } + }, + "environment_name": "BallDrop" +} diff --git a/questions/BallDrop/sample_manifest.json b/questions/BallDrop/sample_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..5c666048a43518dfbc89c57d3c3790334a2f4f9f --- /dev/null +++ b/questions/BallDrop/sample_manifest.json @@ -0,0 +1,2638 @@ +{ + "frost_01234": [ + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "47f84b9f0967540a98380bd737333252", + "b587da7834684a6a9513f7e7b90e8a9e", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "34574575af47569e8859d01ec8078234", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "4373eb3bac935c4e8b80d82e28f50ffe", + "3044f1762eac4229b425c3e811c01448", + "b6f90295752f44209d533b6b652360d6" + ], + "test_samples": [ + "06c7aeb4badb5ca085c44da26f39cfa5", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "34574575af47569e8859d01ec8078234", + "631af851dff9494ca7d21f21dac03a3e", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "831770be078648aebe505cfc3add6f65", + "adfc9499d2b940ef845993221934f87a" + ], + "test_samples_hashes": [ + "9e37b30901ff0d2cafb29885c1d82916", + "08ce3902b9492043d7b112e21f5a3c03", + "619e92593025d6e33838d323838e183b", + "b40c434de3e35557c8d3bf3db99a4066", + "a738a4f972ff9380e15061c67b7b68df", + "bb7ee2cccaf39d16c61581ac97abc3b2", + "ca75e464db9aabb4ec7dd5f7fc14e650", + "098790278431a07c1c909e2692de1562", + "c54666132e0a75eabfa0b838923499e6", + "0a2345ea9190e12b2e5e36189f82e344" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "d72cb57930e8be0d8f09dbae1019b244181e0efd66e9ef6e6c54fc5105f73668", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7df2e9376cf654358d488e51f581b366", + "81cb67a56e3b53dabdc75ab3749d1a65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "b5641f9340fa4cd08749657345a1bedf", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.0, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "81cb67a56e3b53dabdc75ab3749d1a65", + "47f84b9f0967540a98380bd737333252", + "4d36b3bd1e524a49b337e2ac9a97206b", + "63b4f438163057f9b67ff2a42d6427c6", + "7109214388a753679ea3347d46ee75f5", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "b587da7834684a6a9513f7e7b90e8a9e", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "bb0abe5b86f652bf819eb1c65190c4c7", + "4757e7217e7a56d693494ebf2f740d43" + ], + "test_samples": [ + "39766784659f583399ef48bbd86c3a85", + "47f84b9f0967540a98380bd737333252", + "61cd70ea73e1497f981510b030571fe5", + "8f251b7ab3515230926f69d9fcd57dd4", + "93651e54ac445eba83c9f71cd64421a9", + "9583b6b3a5c548c0893c647f3350dd5a", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "c8baff863e805d089948e883182b8d93", + "ccf1034ad54459debd65aff402728604" + ], + "test_samples_hashes": [ + "836e1a8f486930cc0d6a751ca0de8b82", + "9860761527be368b6e6e5315b3fb7a8d", + "4d9199aeba0c49ff1830c314aca3993d", + "45e01075619fadbd7a23034d6ffdc43c", + "7bb74baa7f937506c2dafd80ce10b72e", + "2f867b9ab82db386c070869d95063d53", + "b38204c1c0625735aba0e9b7d875e92f", + "04f9d1b0727010f396fb1684a5c41598", + "326d77c44cb90667e86e8bd4ca6ccae3", + "6e2550ab1675b598fad1d9366529228c" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "efbe2c6821a0b59af799d44d069ea4ca5356457977bbd3a2a632a3de823dd8c2", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "9469a80f76084842933ee512968eadc2", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "2e34f5dce75f46be81d22e0deae7b30f", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "4d36b3bd1e524a49b337e2ac9a97206b", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "4757e7217e7a56d693494ebf2f740d43", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "63b4f438163057f9b67ff2a42d6427c6", + "b6f90295752f44209d533b6b652360d6", + "efeae73360874aefb8d1d01031db168a", + "47f84b9f0967540a98380bd737333252" + ], + "test_samples": [ + "08a23a9f66584605a0ce62ddaf3fcdd8", + "3df1d923d25b44d6826383b65210f7eb", + "4d36b3bd1e524a49b337e2ac9a97206b", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "a236c7b60f6d489cad9fdc8ef847a425", + "cf5492d1ceb754e6af4c83df1d1e646f", + "d72e4ead5b465703b756c36d29213e3f", + "e415e80bfb684ba0a7eec7c7a02ce180", + "efeae73360874aefb8d1d01031db168a", + "fbee8e4e80435cb08b5438b33fb62e36" + ], + "test_samples_hashes": [ + "1d2aa8546042857ed6af7d0cb3c8f281", + "ee64f619e246220eace28b0c6a53043a", + "3071d8a2f38e3bf23b51b9d74fc88ba7", + "2fe840cf37fbab215afa443314133d86", + "dcf375a81cfcfec3202906d88dda5f01", + "32c83999b4b12fc5e691a21ae0123662", + "89c7778dfb923796d6e6ec8ef5758779", + "2a35f33aaeb12783952a0772c11358f7", + "a38a83f2e3d613aafab279e3f271e272", + "2bec3dbcf5c2516b470d8ed570fe42b1" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "4ed450dd5a54875e751918e41943c4ba0a7717bd72b357b8dc1a9b360c914cad", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.1, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "f80e701fc3124cb1bf1cf9a5dfb89792", + "4d36b3bd1e524a49b337e2ac9a97206b", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "b587da7834684a6a9513f7e7b90e8a9e", + "2e34f5dce75f46be81d22e0deae7b30f", + "7df2e9376cf654358d488e51f581b366", + "bb0abe5b86f652bf819eb1c65190c4c7", + "cf90c926e3d14e669525f9577984bec0", + "ea78e1117b7c4e0c8257a07ca20ce3a7" + ], + "test_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "12468fab1955463b813e11f74847029b", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "4c1e9b4e56e32073502d64f3e9b72356", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "bb0abe5b86f652bf819eb1c65190c4c7", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931" + ], + "test_samples_hashes": [ + "008a1a875535610493e9c18fe12514ec", + "f8a979463f10122d61d65d4294e0c4ed", + "e78defd7189961de44bf71c75071bee7", + "39211469f82bdc723cce0f4bb49d586d", + "417f45bfde6f5b38524778418a5c2df9", + "53a21a7cc3cebd2f07f8fc35d9f09b16", + "63f70e3f124cc733cb2ce87a84dece0e", + "64c3874fe139e7f82cea10a84ad55682", + "a8fc50fcb1a8d2d50b71e061d012c60a", + "819df5dc57387bd138a5a6b2f731d574" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "83a62d9c2d795ac172c9a2f9330534b9e95afa2c1c43be23caf6f58cd5aaa439", + "other_samples": [ + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "4373eb3bac935c4e8b80d82e28f50ffe", + "7df2e9376cf654358d488e51f581b366", + "7109214388a753679ea3347d46ee75f5", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "b587da7834684a6a9513f7e7b90e8a9e", + "3044f1762eac4229b425c3e811c01448", + "efeae73360874aefb8d1d01031db168a", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "test_samples": [ + "4373eb3bac935c4e8b80d82e28f50ffe", + "523c9218f52c57068887247f1c7a044a", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "919c229adaf44cd68af7629ae8feb882", + "9469a80f76084842933ee512968eadc2", + "b5641f9340fa4cd08749657345a1bedf", + "e195837b1041539c86a3733e4391198b", + "f4c8f20b5816534db655784eb8d032f6", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "test_samples_hashes": [ + "bbe6a229acb51751e042f219741dd051", + "93eac8ecfdcfad4b83c3c05bb8b69549", + "d75932dfc68110c7928d725d1a6f2b8f", + "63c775e5f88ff8d94dfea59a5b6eb0ea", + "6b8b779f68fdbc2035dac9cc6fe388a9", + "33737f550d7b5bf32a9ca9761e18e797", + "25baa67dc197003efd4ab78cd9496276", + "b382ce19871ef185790c086c584cd601", + "e36e5fb8caecbbb6fa7e910f2a30d7f8", + "88879284370adeb6964750d45f07a613" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "78130909faa5c81019f6b515c46fb0b41b51ce8c1f41448e15f9ac09ddf88199", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + } + } + ], + "fern_01234": [ + { + "train_samples_baselines": [ + "7109214388a753679ea3347d46ee75f5", + "7109214388a753679ea3347d46ee75f5", + "4d36b3bd1e524a49b337e2ac9a97206b", + "4757e7217e7a56d693494ebf2f740d43", + "4757e7217e7a56d693494ebf2f740d43" + ], + "train_samples": [ + "1dcbb406a6365e2f86382f83d5a8a508", + "36aee79d125e52d18d0608ce1873c6b4", + "4d36b3bd1e524a49b337e2ac9a97206b", + "ccf1034ad54459debd65aff402728604", + "e799ab902f045c1d9c29b64ced889454" + ], + "train_samples_hashes": [ + "156430334bacc24f18dce6562ddb8dd0", + "2fa0662ab14d66a321fdab622c17c590", + "3071d8a2f38e3bf23b51b9d74fc88ba7", + "6e2550ab1675b598fad1d9366529228c", + "6e08a6210709db704ec39e8113234e44" + ], + "test_samples_baselines": [ + "47f84b9f0967540a98380bd737333252", + "b587da7834684a6a9513f7e7b90e8a9e", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "34574575af47569e8859d01ec8078234", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "4373eb3bac935c4e8b80d82e28f50ffe", + "3044f1762eac4229b425c3e811c01448", + "b6f90295752f44209d533b6b652360d6" + ], + "test_samples": [ + "06c7aeb4badb5ca085c44da26f39cfa5", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "34574575af47569e8859d01ec8078234", + "631af851dff9494ca7d21f21dac03a3e", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "831770be078648aebe505cfc3add6f65", + "adfc9499d2b940ef845993221934f87a" + ], + "test_samples_hashes": [ + "9e37b30901ff0d2cafb29885c1d82916", + "08ce3902b9492043d7b112e21f5a3c03", + "619e92593025d6e33838d323838e183b", + "b40c434de3e35557c8d3bf3db99a4066", + "a738a4f972ff9380e15061c67b7b68df", + "bb7ee2cccaf39d16c61581ac97abc3b2", + "ca75e464db9aabb4ec7dd5f7fc14e650", + "098790278431a07c1c909e2692de1562", + "c54666132e0a75eabfa0b838923499e6", + "0a2345ea9190e12b2e5e36189f82e344" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "37daf54b39a161a3aa07b50ce8602f3e73c4842d37736f4507c9ce5e32acf031", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "1b4abf9677a956e4bb90051b4494f4b3", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7df2e9376cf654358d488e51f581b366", + "81cb67a56e3b53dabdc75ab3749d1a65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "b5641f9340fa4cd08749657345a1bedf", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.0, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "0d96332e3fae53ff8c19e0445a16f304", + "7df2e9376cf654358d488e51f581b366", + "7df2e9376cf654358d488e51f581b366", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "b6f90295752f44209d533b6b652360d6" + ], + "train_samples": [ + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "c526889cd0b44bdb9efc5223a2d024a6" + ], + "train_samples_hashes": [ + "fec237f2c7761dd4961fab110cf59a36", + "7927f0e02ed2fe57d1885a23b2a4de66", + "63f70e3f124cc733cb2ce87a84dece0e", + "bb7ee2cccaf39d16c61581ac97abc3b2", + "288d2fd25cfba23511dc57230f2d8302" + ], + "test_samples_baselines": [ + "81cb67a56e3b53dabdc75ab3749d1a65", + "47f84b9f0967540a98380bd737333252", + "4d36b3bd1e524a49b337e2ac9a97206b", + "63b4f438163057f9b67ff2a42d6427c6", + "7109214388a753679ea3347d46ee75f5", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "b587da7834684a6a9513f7e7b90e8a9e", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "bb0abe5b86f652bf819eb1c65190c4c7", + "4757e7217e7a56d693494ebf2f740d43" + ], + "test_samples": [ + "39766784659f583399ef48bbd86c3a85", + "47f84b9f0967540a98380bd737333252", + "61cd70ea73e1497f981510b030571fe5", + "8f251b7ab3515230926f69d9fcd57dd4", + "93651e54ac445eba83c9f71cd64421a9", + "9583b6b3a5c548c0893c647f3350dd5a", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "c8baff863e805d089948e883182b8d93", + "ccf1034ad54459debd65aff402728604" + ], + "test_samples_hashes": [ + "836e1a8f486930cc0d6a751ca0de8b82", + "9860761527be368b6e6e5315b3fb7a8d", + "4d9199aeba0c49ff1830c314aca3993d", + "45e01075619fadbd7a23034d6ffdc43c", + "7bb74baa7f937506c2dafd80ce10b72e", + "2f867b9ab82db386c070869d95063d53", + "b38204c1c0625735aba0e9b7d875e92f", + "04f9d1b0727010f396fb1684a5c41598", + "326d77c44cb90667e86e8bd4ca6ccae3", + "6e2550ab1675b598fad1d9366529228c" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "cc4086b420e04ce3fde53470a9af6f7e0aa8ef65beada93471150e145cef0f78", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "9469a80f76084842933ee512968eadc2", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "81cb67a56e3b53dabdc75ab3749d1a65", + "fca8a26ed97e46f5a04e19c22db27fa5", + "4373eb3bac935c4e8b80d82e28f50ffe", + "b587da7834684a6a9513f7e7b90e8a9e", + "7109214388a753679ea3347d46ee75f5" + ], + "train_samples": [ + "39766784659f583399ef48bbd86c3a85", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "4373eb3bac935c4e8b80d82e28f50ffe", + "919c229adaf44cd68af7629ae8feb882", + "9c445001c2f75ddb9bd1d47993a664f2" + ], + "train_samples_hashes": [ + "836e1a8f486930cc0d6a751ca0de8b82", + "18f315511ca399ede8d2a8bd7ca39601", + "bbe6a229acb51751e042f219741dd051", + "6b8b779f68fdbc2035dac9cc6fe388a9", + "c3fc753c70d195ba992e129fd1bfe55a" + ], + "test_samples_baselines": [ + "2e34f5dce75f46be81d22e0deae7b30f", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "4d36b3bd1e524a49b337e2ac9a97206b", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "4757e7217e7a56d693494ebf2f740d43", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "63b4f438163057f9b67ff2a42d6427c6", + "b6f90295752f44209d533b6b652360d6", + "efeae73360874aefb8d1d01031db168a", + "47f84b9f0967540a98380bd737333252" + ], + "test_samples": [ + "08a23a9f66584605a0ce62ddaf3fcdd8", + "3df1d923d25b44d6826383b65210f7eb", + "4d36b3bd1e524a49b337e2ac9a97206b", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "a236c7b60f6d489cad9fdc8ef847a425", + "cf5492d1ceb754e6af4c83df1d1e646f", + "d72e4ead5b465703b756c36d29213e3f", + "e415e80bfb684ba0a7eec7c7a02ce180", + "efeae73360874aefb8d1d01031db168a", + "fbee8e4e80435cb08b5438b33fb62e36" + ], + "test_samples_hashes": [ + "1d2aa8546042857ed6af7d0cb3c8f281", + "ee64f619e246220eace28b0c6a53043a", + "3071d8a2f38e3bf23b51b9d74fc88ba7", + "2fe840cf37fbab215afa443314133d86", + "dcf375a81cfcfec3202906d88dda5f01", + "32c83999b4b12fc5e691a21ae0123662", + "89c7778dfb923796d6e6ec8ef5758779", + "2a35f33aaeb12783952a0772c11358f7", + "a38a83f2e3d613aafab279e3f271e272", + "2bec3dbcf5c2516b470d8ed570fe42b1" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "7b6e88464e899753dc56ed1bdd5f533d36fdbe43782068550b68f5b9f8f164c8", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9583b6b3a5c548c0893c647f3350dd5a", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.1, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "34574575af47569e8859d01ec8078234", + "0f159f29c1004403a466c6bea03d7d02", + "efeae73360874aefb8d1d01031db168a", + "efeae73360874aefb8d1d01031db168a", + "4373eb3bac935c4e8b80d82e28f50ffe" + ], + "train_samples": [ + "34574575af47569e8859d01ec8078234", + "8845ac4101424920bfad9f1eebc21c1c", + "a1c583110649c9166d2b8cc4ab82eca6", + "ab2618c21bea42a5b9d58b93173c082c", + "cd023a4ad4de4316bba0e94b5bf35d8c" + ], + "train_samples_hashes": [ + "b40c434de3e35557c8d3bf3db99a4066", + "5cabff5e2d8d11c28d26f779785f0f08", + "88f04e706313e248a820796b502a1348", + "dc6600b666dd665684df177b2470b81f", + "27dbdcaaa6d7b3190901c22e0f90ce88" + ], + "test_samples_baselines": [ + "f80e701fc3124cb1bf1cf9a5dfb89792", + "4d36b3bd1e524a49b337e2ac9a97206b", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "b587da7834684a6a9513f7e7b90e8a9e", + "2e34f5dce75f46be81d22e0deae7b30f", + "7df2e9376cf654358d488e51f581b366", + "bb0abe5b86f652bf819eb1c65190c4c7", + "cf90c926e3d14e669525f9577984bec0", + "ea78e1117b7c4e0c8257a07ca20ce3a7" + ], + "test_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "12468fab1955463b813e11f74847029b", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "4c1e9b4e56e32073502d64f3e9b72356", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "bb0abe5b86f652bf819eb1c65190c4c7", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931" + ], + "test_samples_hashes": [ + "008a1a875535610493e9c18fe12514ec", + "f8a979463f10122d61d65d4294e0c4ed", + "e78defd7189961de44bf71c75071bee7", + "39211469f82bdc723cce0f4bb49d586d", + "417f45bfde6f5b38524778418a5c2df9", + "53a21a7cc3cebd2f07f8fc35d9f09b16", + "63f70e3f124cc733cb2ce87a84dece0e", + "64c3874fe139e7f82cea10a84ad55682", + "a8fc50fcb1a8d2d50b71e061d012c60a", + "819df5dc57387bd138a5a6b2f731d574" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "e602de54114b9722633ceb57592b1a1470ed0793432fd3c586d9d3e411da0ce5", + "other_samples": [ + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cf5492d1ceb754e6af4c83df1d1e646f", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "2e34f5dce75f46be81d22e0deae7b30f", + "63b4f438163057f9b67ff2a42d6427c6", + "81cb67a56e3b53dabdc75ab3749d1a65", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "63b4f438163057f9b67ff2a42d6427c6" + ], + "train_samples": [ + "08a23a9f66584605a0ce62ddaf3fcdd8", + "63b4f438163057f9b67ff2a42d6427c6", + "6c7ac291a99852eea1bd5fb4a619a097", + "a58f39410c8a58289fe81de42778819a", + "f73e9eeb29be5e07b8c13705ff2ab2a6" + ], + "train_samples_hashes": [ + "1d2aa8546042857ed6af7d0cb3c8f281", + "2a7ceb16692073467365faa0c1cbd44a", + "ac3bba2bc1090b7838d5e1e47312d9f5", + "fa54d7d0b0be210b7c47e2a8a8a777a3", + "77dbaedb84822756a608e441ea18ff96" + ], + "test_samples_baselines": [ + "4373eb3bac935c4e8b80d82e28f50ffe", + "7df2e9376cf654358d488e51f581b366", + "7109214388a753679ea3347d46ee75f5", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "b587da7834684a6a9513f7e7b90e8a9e", + "3044f1762eac4229b425c3e811c01448", + "efeae73360874aefb8d1d01031db168a", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "test_samples": [ + "4373eb3bac935c4e8b80d82e28f50ffe", + "523c9218f52c57068887247f1c7a044a", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "919c229adaf44cd68af7629ae8feb882", + "9469a80f76084842933ee512968eadc2", + "b5641f9340fa4cd08749657345a1bedf", + "e195837b1041539c86a3733e4391198b", + "f4c8f20b5816534db655784eb8d032f6", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "test_samples_hashes": [ + "bbe6a229acb51751e042f219741dd051", + "93eac8ecfdcfad4b83c3c05bb8b69549", + "d75932dfc68110c7928d725d1a6f2b8f", + "63c775e5f88ff8d94dfea59a5b6eb0ea", + "6b8b779f68fdbc2035dac9cc6fe388a9", + "33737f550d7b5bf32a9ca9761e18e797", + "25baa67dc197003efd4ab78cd9496276", + "b382ce19871ef185790c086c584cd601", + "e36e5fb8caecbbb6fa7e910f2a30d7f8", + "88879284370adeb6964750d45f07a613" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "03809ec07caa89653526f89ff39f741d7359a485cc91619bba8ec54e45369e54", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + } + } + ], + "gentle_01234": [ + { + "train_samples_baselines": [ + "0d96332e3fae53ff8c19e0445a16f304", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e34f5dce75f46be81d22e0deae7b30f", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "4757e7217e7a56d693494ebf2f740d43", + "7109214388a753679ea3347d46ee75f5", + "81cb67a56e3b53dabdc75ab3749d1a65", + "63b4f438163057f9b67ff2a42d6427c6", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "cf90c926e3d14e669525f9577984bec0", + "efeae73360874aefb8d1d01031db168a", + "efeae73360874aefb8d1d01031db168a", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "63b4f438163057f9b67ff2a42d6427c6", + "7109214388a753679ea3347d46ee75f5" + ], + "train_samples": [ + "0d96332e3fae53ff8c19e0445a16f304", + "2e34f5dce75f46be81d22e0deae7b30f", + "39330e4b3b284bfa859687cd7abf9b14", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "74558dec52105edfb945508fea345732", + "81cb67a56e3b53dabdc75ab3749d1a65", + "8f251b7ab3515230926f69d9fcd57dd4", + "9583b6b3a5c548c0893c647f3350dd5a", + "a26263883fa446e49b5b921cfdf3944d", + "ab2618c21bea42a5b9d58b93173c082c", + "b5641f9340fa4cd08749657345a1bedf", + "cf5492d1ceb754e6af4c83df1d1e646f", + "d72e4ead5b465703b756c36d29213e3f", + "fbe454d4b3625caaaa07aaed18d16268" + ], + "train_samples_hashes": [ + "07cb1ec08998ac5324c96499966d920b", + "43c806fe8735b2e4aad52cb1ebc76c9b", + "2d43c540013e4b679cbd6eead7f84c80", + "6b516d1358c9c9ad27e34c5e0fb2030b", + "22685ffed6e8d27bcaa6c0b78e351417", + "d75932dfc68110c7928d725d1a6f2b8f", + "3b5055469dd37e90381e8b7a519b8308", + "45e01075619fadbd7a23034d6ffdc43c", + "2f867b9ab82db386c070869d95063d53", + "bf3087c0d4645277b415dc4e9dcdb788", + "dc6600b666dd665684df177b2470b81f", + "25baa67dc197003efd4ab78cd9496276", + "32c83999b4b12fc5e691a21ae0123662", + "89c7778dfb923796d6e6ec8ef5758779", + "664e12e057bc5745ba8b6a0b17907b2e" + ], + "test_samples_baselines": [ + "47f84b9f0967540a98380bd737333252", + "b587da7834684a6a9513f7e7b90e8a9e", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "34574575af47569e8859d01ec8078234", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "4373eb3bac935c4e8b80d82e28f50ffe", + "3044f1762eac4229b425c3e811c01448", + "b6f90295752f44209d533b6b652360d6" + ], + "test_samples": [ + "06c7aeb4badb5ca085c44da26f39cfa5", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "34574575af47569e8859d01ec8078234", + "631af851dff9494ca7d21f21dac03a3e", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "831770be078648aebe505cfc3add6f65", + "adfc9499d2b940ef845993221934f87a" + ], + "test_samples_hashes": [ + "9e37b30901ff0d2cafb29885c1d82916", + "08ce3902b9492043d7b112e21f5a3c03", + "619e92593025d6e33838d323838e183b", + "b40c434de3e35557c8d3bf3db99a4066", + "a738a4f972ff9380e15061c67b7b68df", + "bb7ee2cccaf39d16c61581ac97abc3b2", + "ca75e464db9aabb4ec7dd5f7fc14e650", + "098790278431a07c1c909e2692de1562", + "c54666132e0a75eabfa0b838923499e6", + "0a2345ea9190e12b2e5e36189f82e344" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "b339fec75bc5ee04724ae000b411dad4dc0ca43ecfe6282e187886b7526c1859", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "36aee79d125e52d18d0608ce1873c6b4", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7df2e9376cf654358d488e51f581b366", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.0, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "3044f1762eac4229b425c3e811c01448", + "efeae73360874aefb8d1d01031db168a", + "3044f1762eac4229b425c3e811c01448", + "34574575af47569e8859d01ec8078234", + "7df2e9376cf654358d488e51f581b366", + "0f159f29c1004403a466c6bea03d7d02", + "0d96332e3fae53ff8c19e0445a16f304", + "efeae73360874aefb8d1d01031db168a", + "b6f90295752f44209d533b6b652360d6", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "b6f90295752f44209d533b6b652360d6", + "34574575af47569e8859d01ec8078234", + "b6f90295752f44209d533b6b652360d6", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "f80e701fc3124cb1bf1cf9a5dfb89792" + ], + "train_samples": [ + "2711a07092314ea681315ca82e5bc51b", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "42e0669504784adea7e79a929f5c5495", + "55cd4372b9205e238624bf60157b1250", + "7df2e9376cf654358d488e51f581b366", + "9d38efbdeee64affb143c411f6081d62", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "adfc9499d2b940ef845993221934f87a", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "ca218ab4f28747f6b38fc0149e36eb9b", + "d5e619d9d7fb52898940b8140f653c12", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "f80e701fc3124cb1bf1cf9a5dfb89792" + ], + "train_samples_hashes": [ + "047ddfaf94e03e7940e09b1ad70616ab", + "4aeb4e670f05f569f0f8133a3cfedd77", + "1c6208a72d86d3824f214c3a12ed1468", + "ae7ad5d062eeca72240280476370ae0b", + "8ee58fd5b895962534f77d7c538d6b67", + "2382887fb18a1662b1f5539350b236af", + "d7018f58cb2a1b9def853739f1fc7f9d", + "88f04e706313e248a820796b502a1348", + "0a2345ea9190e12b2e5e36189f82e344", + "9bec6d5ac5e767ece2dddf6301cbb7e3", + "c150a28c0f7311280884ac76e0927c20", + "a0e3858acd774f53864295bc32fb89ca", + "2a35f33aaeb12783952a0772c11358f7", + "bbeec99eec6011df3dd9567215e34028", + "5be979d33412c625f2fe6d0ba1739af4" + ], + "test_samples_baselines": [ + "81cb67a56e3b53dabdc75ab3749d1a65", + "47f84b9f0967540a98380bd737333252", + "4d36b3bd1e524a49b337e2ac9a97206b", + "63b4f438163057f9b67ff2a42d6427c6", + "7109214388a753679ea3347d46ee75f5", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "b587da7834684a6a9513f7e7b90e8a9e", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "bb0abe5b86f652bf819eb1c65190c4c7", + "4757e7217e7a56d693494ebf2f740d43" + ], + "test_samples": [ + "39766784659f583399ef48bbd86c3a85", + "47f84b9f0967540a98380bd737333252", + "61cd70ea73e1497f981510b030571fe5", + "8f251b7ab3515230926f69d9fcd57dd4", + "93651e54ac445eba83c9f71cd64421a9", + "9583b6b3a5c548c0893c647f3350dd5a", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "c8baff863e805d089948e883182b8d93", + "ccf1034ad54459debd65aff402728604" + ], + "test_samples_hashes": [ + "836e1a8f486930cc0d6a751ca0de8b82", + "9860761527be368b6e6e5315b3fb7a8d", + "4d9199aeba0c49ff1830c314aca3993d", + "45e01075619fadbd7a23034d6ffdc43c", + "7bb74baa7f937506c2dafd80ce10b72e", + "2f867b9ab82db386c070869d95063d53", + "b38204c1c0625735aba0e9b7d875e92f", + "04f9d1b0727010f396fb1684a5c41598", + "326d77c44cb90667e86e8bd4ca6ccae3", + "6e2550ab1675b598fad1d9366529228c" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "17756883d39570e7b0565b3bb72f7e6cef4635d36106fcda0bfdbc16f584cae1", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "9469a80f76084842933ee512968eadc2", + "9c445001c2f75ddb9bd1d47993a664f2", + "9daf764b6051540e98b175ec43c551c4", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "b5641f9340fa4cd08749657345a1bedf", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "ca1727e8ff895ff497d01835818489e8", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "0f159f29c1004403a466c6bea03d7d02", + "b587da7834684a6a9513f7e7b90e8a9e", + "0d96332e3fae53ff8c19e0445a16f304", + "4373eb3bac935c4e8b80d82e28f50ffe", + "0d96332e3fae53ff8c19e0445a16f304", + "7df2e9376cf654358d488e51f581b366", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "4373eb3bac935c4e8b80d82e28f50ffe", + "0f159f29c1004403a466c6bea03d7d02", + "cf90c926e3d14e669525f9577984bec0", + "3044f1762eac4229b425c3e811c01448", + "b587da7834684a6a9513f7e7b90e8a9e", + "4373eb3bac935c4e8b80d82e28f50ffe", + "cf90c926e3d14e669525f9577984bec0", + "ba2e3cbfe88c52748c6d2c81c7a04600" + ], + "train_samples": [ + "0f159f29c1004403a466c6bea03d7d02", + "2348ceba66234f5f91acc812e273a3d5", + "303d59b812a942eeac4a05f71df85ff1", + "3e706177423c5f289117291cc571bc8d", + "50b90a6df0e25ac480a3af80bb1b21bd", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "77968ae18bca566cbcffba3830a3fece", + "81867907eb7c5370b8f73d2718d8398d", + "8845ac4101424920bfad9f1eebc21c1c", + "8f30e32b7b6b4d04915b710074abb452", + "9469a80f76084842933ee512968eadc2", + "b587da7834684a6a9513f7e7b90e8a9e", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf90c926e3d14e669525f9577984bec0", + "e195837b1041539c86a3733e4391198b" + ], + "train_samples_hashes": [ + "81edab40626ea7642f819abeeedf80c7", + "417f45bfde6f5b38524778418a5c2df9", + "fec237f2c7761dd4961fab110cf59a36", + "d03490834b5ae6e36583fe7b0b6c0cb5", + "a390bd7b013aeba50d900129c8632ed0", + "63f70e3f124cc733cb2ce87a84dece0e", + "1556a1f828847ba49d9ade99e0c2de54", + "098790278431a07c1c909e2692de1562", + "5cabff5e2d8d11c28d26f779785f0f08", + "565f59a9c3a6aa158a5f44abda8371a6", + "33737f550d7b5bf32a9ca9761e18e797", + "b38204c1c0625735aba0e9b7d875e92f", + "27dbdcaaa6d7b3190901c22e0f90ce88", + "a8fc50fcb1a8d2d50b71e061d012c60a", + "b382ce19871ef185790c086c584cd601" + ], + "test_samples_baselines": [ + "2e34f5dce75f46be81d22e0deae7b30f", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "4d36b3bd1e524a49b337e2ac9a97206b", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "4757e7217e7a56d693494ebf2f740d43", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "63b4f438163057f9b67ff2a42d6427c6", + "b6f90295752f44209d533b6b652360d6", + "efeae73360874aefb8d1d01031db168a", + "47f84b9f0967540a98380bd737333252" + ], + "test_samples": [ + "08a23a9f66584605a0ce62ddaf3fcdd8", + "3df1d923d25b44d6826383b65210f7eb", + "4d36b3bd1e524a49b337e2ac9a97206b", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "a236c7b60f6d489cad9fdc8ef847a425", + "cf5492d1ceb754e6af4c83df1d1e646f", + "d72e4ead5b465703b756c36d29213e3f", + "e415e80bfb684ba0a7eec7c7a02ce180", + "efeae73360874aefb8d1d01031db168a", + "fbee8e4e80435cb08b5438b33fb62e36" + ], + "test_samples_hashes": [ + "1d2aa8546042857ed6af7d0cb3c8f281", + "ee64f619e246220eace28b0c6a53043a", + "3071d8a2f38e3bf23b51b9d74fc88ba7", + "2fe840cf37fbab215afa443314133d86", + "dcf375a81cfcfec3202906d88dda5f01", + "32c83999b4b12fc5e691a21ae0123662", + "89c7778dfb923796d6e6ec8ef5758779", + "2a35f33aaeb12783952a0772c11358f7", + "a38a83f2e3d613aafab279e3f271e272", + "2bec3dbcf5c2516b470d8ed570fe42b1" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "35387e92771a8892c43272680404c5461093fc52c58e374a68799a088b27f54e", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "0d96332e3fae53ff8c19e0445a16f304", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4c1e9b4e56e32073502d64f3e9b72356", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8f251b7ab3515230926f69d9fcd57dd4", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.1, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "4373eb3bac935c4e8b80d82e28f50ffe", + "3c523b45ab52505d8d9b7c74e5ec21e9", + "34574575af47569e8859d01ec8078234", + "7109214388a753679ea3347d46ee75f5", + "81cb67a56e3b53dabdc75ab3749d1a65", + "fca8a26ed97e46f5a04e19c22db27fa5", + "63b4f438163057f9b67ff2a42d6427c6", + "63b4f438163057f9b67ff2a42d6427c6", + "7109214388a753679ea3347d46ee75f5", + "4373eb3bac935c4e8b80d82e28f50ffe", + "b6f90295752f44209d533b6b652360d6", + "63b4f438163057f9b67ff2a42d6427c6", + "fca8a26ed97e46f5a04e19c22db27fa5", + "4757e7217e7a56d693494ebf2f740d43", + "efeae73360874aefb8d1d01031db168a" + ], + "train_samples": [ + "1b4abf9677a956e4bb90051b4494f4b3", + "2e86956a56055b2aa94c0030af2f2f74", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39766784659f583399ef48bbd86c3a85", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "56329b0d596f5b4daa3db101e5a634e6", + "63b4f438163057f9b67ff2a42d6427c6", + "9c445001c2f75ddb9bd1d47993a664f2", + "a7a514d5d4565e85b4f48ad78735021e", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a" + ], + "train_samples_hashes": [ + "d0f642a5f67ba8ba03c2e73321c87bd0", + "55f8d69efb3cee7479c9f93cfebf23d5", + "b40c434de3e35557c8d3bf3db99a4066", + "2fa0662ab14d66a321fdab622c17c590", + "836e1a8f486930cc0d6a751ca0de8b82", + "18f315511ca399ede8d2a8bd7ca39601", + "b67fddc78fbf439909b7f85258550959", + "2a7ceb16692073467365faa0c1cbd44a", + "c3fc753c70d195ba992e129fd1bfe55a", + "05b4ad4ca7a6cb1b75aefd36e6c87267", + "288d2fd25cfba23511dc57230f2d8302", + "4244ddfb23d3513ea6438f4ab22dfb58", + "0ca42568cddbf4e65e627772ee841644", + "6e08a6210709db704ec39e8113234e44", + "a38a83f2e3d613aafab279e3f271e272" + ], + "test_samples_baselines": [ + "f80e701fc3124cb1bf1cf9a5dfb89792", + "4d36b3bd1e524a49b337e2ac9a97206b", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "b587da7834684a6a9513f7e7b90e8a9e", + "2e34f5dce75f46be81d22e0deae7b30f", + "7df2e9376cf654358d488e51f581b366", + "bb0abe5b86f652bf819eb1c65190c4c7", + "cf90c926e3d14e669525f9577984bec0", + "ea78e1117b7c4e0c8257a07ca20ce3a7" + ], + "test_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "12468fab1955463b813e11f74847029b", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "4c1e9b4e56e32073502d64f3e9b72356", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "bb0abe5b86f652bf819eb1c65190c4c7", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931" + ], + "test_samples_hashes": [ + "008a1a875535610493e9c18fe12514ec", + "f8a979463f10122d61d65d4294e0c4ed", + "e78defd7189961de44bf71c75071bee7", + "39211469f82bdc723cce0f4bb49d586d", + "417f45bfde6f5b38524778418a5c2df9", + "53a21a7cc3cebd2f07f8fc35d9f09b16", + "63f70e3f124cc733cb2ce87a84dece0e", + "64c3874fe139e7f82cea10a84ad55682", + "a8fc50fcb1a8d2d50b71e061d012c60a", + "819df5dc57387bd138a5a6b2f731d574" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "287a50567f251427ca23137233fb5a606d0f79c0e0f0bbb9c1b54bbc2f211287", + "other_samples": [ + "00f96c534483544fbf5cb3d14e38272f", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1dcbb406a6365e2f86382f83d5a8a508", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "39330e4b3b284bfa859687cd7abf9b14", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4373eb3bac935c4e8b80d82e28f50ffe", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "523c9218f52c57068887247f1c7a044a", + "55cd4372b9205e238624bf60157b1250", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "6228a7c919375242b6065fac22112b1e", + "631af851dff9494ca7d21f21dac03a3e", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "77968ae18bca566cbcffba3830a3fece", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "873c2abc53235d5eb408430e7313c0b7", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "919c229adaf44cd68af7629ae8feb882", + "91ddb1413c4fbf36279927c5338fc650", + "93651e54ac445eba83c9f71cd64421a9", + "9469a80f76084842933ee512968eadc2", + "9583b6b3a5c548c0893c647f3350dd5a", + "9d38efbdeee64affb143c411f6081d62", + "9daf764b6051540e98b175ec43c551c4", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a6e56514770442afb6af22c381df3a0a", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b5641f9340fa4cd08749657345a1bedf", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "c8baff863e805d089948e883182b8d93", + "ca1727e8ff895ff497d01835818489e8", + "ca218ab4f28747f6b38fc0149e36eb9b", + "ccf1034ad54459debd65aff402728604", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "dcbbcb9aaba84f95b6e6c410a573039a", + "e195837b1041539c86a3733e4391198b", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "f4c8f20b5816534db655784eb8d032f6", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "0d96332e3fae53ff8c19e0445a16f304", + "b6f90295752f44209d533b6b652360d6", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "47f84b9f0967540a98380bd737333252", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "63b4f438163057f9b67ff2a42d6427c6", + "0f159f29c1004403a466c6bea03d7d02", + "34574575af47569e8859d01ec8078234", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "0f159f29c1004403a466c6bea03d7d02", + "4757e7217e7a56d693494ebf2f740d43", + "34574575af47569e8859d01ec8078234", + "4757e7217e7a56d693494ebf2f740d43", + "7d1b3f1a85c049ff9b8bd7c7a18d4041" + ], + "train_samples": [ + "00f96c534483544fbf5cb3d14e38272f", + "439d72f79dac13bab27ea2190255b070", + "4757e7217e7a56d693494ebf2f740d43", + "47f84b9f0967540a98380bd737333252", + "6228a7c919375242b6065fac22112b1e", + "7d1b3f1a85c049ff9b8bd7c7a18d4041", + "873c2abc53235d5eb408430e7313c0b7", + "91ddb1413c4fbf36279927c5338fc650", + "9daf764b6051540e98b175ec43c551c4", + "a6e56514770442afb6af22c381df3a0a", + "bc7a8ee50c0b4cd3bf478e3b3f4230dd", + "c3eb6b6c9abe403faec7dd1a28b4378e", + "ca1727e8ff895ff497d01835818489e8", + "ccf1034ad54459debd65aff402728604", + "dcbbcb9aaba84f95b6e6c410a573039a" + ], + "train_samples_hashes": [ + "38aa0b9033d2e8106cfcbf264f4825e9", + "3466725c4ee603ab1f14711b20fa9779", + "a58cb5cd7bb9b9e6ced5502311b0b7fa", + "9860761527be368b6e6e5315b3fb7a8d", + "2bd1aab36ab5b595de55d70fad4b0509", + "bb7ee2cccaf39d16c61581ac97abc3b2", + "044d90f79212984839c168124645adf7", + "4274cf48806e11de56f279d9e112f5e3", + "104c2d85c6ad7c82eb0d205855155b1b", + "95b230d6e9a56e6ff38939b7531573ef", + "6891d39b1dff50b72cbb305d4285b096", + "b29811a5b1b78bf2be855fe3cae949bb", + "db1b00130d23124ded6b33056368abf6", + "6e2550ab1675b598fad1d9366529228c", + "c0595eea074c071e3ea0d12bf5cb386f" + ], + "test_samples_baselines": [ + "4373eb3bac935c4e8b80d82e28f50ffe", + "7df2e9376cf654358d488e51f581b366", + "7109214388a753679ea3347d46ee75f5", + "ea78e1117b7c4e0c8257a07ca20ce3a7", + "b587da7834684a6a9513f7e7b90e8a9e", + "3044f1762eac4229b425c3e811c01448", + "efeae73360874aefb8d1d01031db168a", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "test_samples": [ + "4373eb3bac935c4e8b80d82e28f50ffe", + "523c9218f52c57068887247f1c7a044a", + "74558dec52105edfb945508fea345732", + "768257b526264466aff05fd9c41ab915", + "919c229adaf44cd68af7629ae8feb882", + "9469a80f76084842933ee512968eadc2", + "b5641f9340fa4cd08749657345a1bedf", + "e195837b1041539c86a3733e4391198b", + "f4c8f20b5816534db655784eb8d032f6", + "fca8a26ed97e46f5a04e19c22db27fa5" + ], + "test_samples_hashes": [ + "bbe6a229acb51751e042f219741dd051", + "93eac8ecfdcfad4b83c3c05bb8b69549", + "d75932dfc68110c7928d725d1a6f2b8f", + "63c775e5f88ff8d94dfea59a5b6eb0ea", + "6b8b779f68fdbc2035dac9cc6fe388a9", + "33737f550d7b5bf32a9ca9761e18e797", + "25baa67dc197003efd4ab78cd9496276", + "b382ce19871ef185790c086c584cd601", + "e36e5fb8caecbbb6fa7e910f2a30d7f8", + "88879284370adeb6964750d45f07a613" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "79e93bab7a0362f6ebf9f8d9d529ef9d7093cee1ff9e1ea9b6499a56ced44d98", + "other_samples": [ + "001063c5dfa248dbbba122ff6743af5f", + "00ae112c60ec4dd1acd1bacbec2560c9", + "06c7aeb4badb5ca085c44da26f39cfa5", + "08a23a9f66584605a0ce62ddaf3fcdd8", + "0d96332e3fae53ff8c19e0445a16f304", + "0f159f29c1004403a466c6bea03d7d02", + "12468fab1955463b813e11f74847029b", + "17a033a89614421b8bb520c3bb286b60", + "18e695472b2d52058cf6deee57c49cf8", + "1b4abf9677a956e4bb90051b4494f4b3", + "1dcbb406a6365e2f86382f83d5a8a508", + "1f176712882c5ce69f5f87c4473b6694", + "2348ceba66234f5f91acc812e273a3d5", + "2711a07092314ea681315ca82e5bc51b", + "2e34f5dce75f46be81d22e0deae7b30f", + "2e86956a56055b2aa94c0030af2f2f74", + "303d59b812a942eeac4a05f71df85ff1", + "33508fdf1d3d5b62a0f61a2ab5200441", + "34574575af47569e8859d01ec8078234", + "36aee79d125e52d18d0608ce1873c6b4", + "39330e4b3b284bfa859687cd7abf9b14", + "39766784659f583399ef48bbd86c3a85", + "3a5ad8a0c7a84fd3a5baaec0eee7a126", + "3df1d923d25b44d6826383b65210f7eb", + "3e28f2d025fd4f229ce62a07f1ed72eb", + "3e706177423c5f289117291cc571bc8d", + "42e0669504784adea7e79a929f5c5495", + "4c1e9b4e56e32073502d64f3e9b72356", + "4d36b3bd1e524a49b337e2ac9a97206b", + "50b90a6df0e25ac480a3af80bb1b21bd", + "55cd4372b9205e238624bf60157b1250", + "56329b0d596f5b4daa3db101e5a634e6", + "568b3e9ed49658d7abaae622daac8cef", + "5f875cc969b55f368b95aee1a44d01e5", + "61cd70ea73e1497f981510b030571fe5", + "631af851dff9494ca7d21f21dac03a3e", + "63b4f438163057f9b67ff2a42d6427c6", + "6514e4c5d16f5fac9810e4b4e0db5d24", + "6c0a4bc64f2f4ff99f76cd41e1ab474d", + "6c7ac291a99852eea1bd5fb4a619a097", + "77968ae18bca566cbcffba3830a3fece", + "7df2e9376cf654358d488e51f581b366", + "7e0f85ac769c5ab78201f4cdac386d36", + "81867907eb7c5370b8f73d2718d8398d", + "81cb67a56e3b53dabdc75ab3749d1a65", + "831770be078648aebe505cfc3add6f65", + "8845ac4101424920bfad9f1eebc21c1c", + "8f251b7ab3515230926f69d9fcd57dd4", + "8f30e32b7b6b4d04915b710074abb452", + "93651e54ac445eba83c9f71cd64421a9", + "9583b6b3a5c548c0893c647f3350dd5a", + "9c445001c2f75ddb9bd1d47993a664f2", + "9d38efbdeee64affb143c411f6081d62", + "9efdb21c07b5589b9b02b808cfd7d78f", + "a1c583110649c9166d2b8cc4ab82eca6", + "a236c7b60f6d489cad9fdc8ef847a425", + "a26263883fa446e49b5b921cfdf3944d", + "a58f39410c8a58289fe81de42778819a", + "a7a514d5d4565e85b4f48ad78735021e", + "ab2618c21bea42a5b9d58b93173c082c", + "adfc9499d2b940ef845993221934f87a", + "b587da7834684a6a9513f7e7b90e8a9e", + "b7c15bc7416f4169b9f2a7166c4865e3", + "ba2e3cbfe88c52748c6d2c81c7a04600", + "bb0abe5b86f652bf819eb1c65190c4c7", + "c526889cd0b44bdb9efc5223a2d024a6", + "c5333739516d5fbfb8aa22aaf11a7cc9", + "c8baff863e805d089948e883182b8d93", + "ca218ab4f28747f6b38fc0149e36eb9b", + "cd023a4ad4de4316bba0e94b5bf35d8c", + "cf5492d1ceb754e6af4c83df1d1e646f", + "cf90c926e3d14e669525f9577984bec0", + "cfc1dde0ab3fe2803be3d63dce48d931", + "d5e208a1a02e48d4ba8d47f9cf10a514", + "d5e619d9d7fb52898940b8140f653c12", + "d72e4ead5b465703b756c36d29213e3f", + "e415e80bfb684ba0a7eec7c7a02ce180", + "e57e705e527d514db713ff28280d11fd", + "e799ab902f045c1d9c29b64ced889454", + "efeae73360874aefb8d1d01031db168a", + "f73e9eeb29be5e07b8c13705ff2ab2a6", + "f80e701fc3124cb1bf1cf9a5dfb89792", + "fbe454d4b3625caaaa07aaed18d16268", + "fbee8e4e80435cb08b5438b33fb62e36" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + } + } + ] +} \ No newline at end of file diff --git a/questions/BounceBall/dataframes/0272b7952029539d9bd40ce24d621991.parquet b/questions/BounceBall/dataframes/0272b7952029539d9bd40ce24d621991.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3dfbe81557142db9a88f827137fbbf5ea5031f9b --- /dev/null +++ b/questions/BounceBall/dataframes/0272b7952029539d9bd40ce24d621991.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5735d85a7ec806c91e428d6a36a9b74f395832a4fd61dc82b7b3f4274d053fe +size 52506 diff --git a/questions/BounceBall/dataframes/0310deb38f465db18ad47c098129e23d.parquet b/questions/BounceBall/dataframes/0310deb38f465db18ad47c098129e23d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ede76cb9a1075c2f087115d47fa8e73e7ce212ac --- /dev/null +++ b/questions/BounceBall/dataframes/0310deb38f465db18ad47c098129e23d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58bf3e985bcae549de0d1bcf7715cddf899c89a63ba325950a53a01aacf23e7b +size 52510 diff --git a/questions/BounceBall/dataframes/062a37da2b3c523ea65fe50198329901.parquet b/questions/BounceBall/dataframes/062a37da2b3c523ea65fe50198329901.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ee57295e414fdf5c6dbdd77af72f9da25ed0d030 --- /dev/null +++ b/questions/BounceBall/dataframes/062a37da2b3c523ea65fe50198329901.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e6f4aa030e1b9c9a7478500b5737447c60883b2106918c2d5c56f344182dec0 +size 52510 diff --git a/questions/BounceBall/dataframes/07f55a16d69449fcb096b003f6593691.parquet b/questions/BounceBall/dataframes/07f55a16d69449fcb096b003f6593691.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c96b8d5ef67803c1503db5526507e6ef48df14ac --- /dev/null +++ b/questions/BounceBall/dataframes/07f55a16d69449fcb096b003f6593691.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d035bd1aa1d85d874b3e2b909797473ca24b94dab5996491836bdf7d68d645ef +size 52545 diff --git a/questions/BounceBall/dataframes/0dda565c44d35ddc944cae93e806df1d.parquet b/questions/BounceBall/dataframes/0dda565c44d35ddc944cae93e806df1d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cab82f198f2007b8332b9cc7eb4a55b88f57cd83 --- /dev/null +++ b/questions/BounceBall/dataframes/0dda565c44d35ddc944cae93e806df1d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68d59dc816c7a54e7d76be77712d1d140b6e0eafd18a3ee38bae3a81ef448dbf +size 52520 diff --git a/questions/BounceBall/dataframes/1032bb73c0b55d1ba0c4685bb20f08f6.parquet b/questions/BounceBall/dataframes/1032bb73c0b55d1ba0c4685bb20f08f6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c8d360cd838c06270335c083333fa3e88da9b168 --- /dev/null +++ b/questions/BounceBall/dataframes/1032bb73c0b55d1ba0c4685bb20f08f6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5cf6297ee1d4f9fe6507beee84823020053247da1965bee5628c2b6f64b6eea +size 52534 diff --git a/questions/BounceBall/dataframes/105dab6037814698b8c5fb1b7cb0d358.parquet b/questions/BounceBall/dataframes/105dab6037814698b8c5fb1b7cb0d358.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ff6f53f1db152739994264cecf3495a0358f3900 --- /dev/null +++ b/questions/BounceBall/dataframes/105dab6037814698b8c5fb1b7cb0d358.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cbe7ad6ffa123a085d6ddf1e0b9b67a227f18fa127cf5b333550d203b08c6b5b +size 52534 diff --git a/questions/BounceBall/dataframes/105de8c7a17f4728931344dd36a28313.parquet b/questions/BounceBall/dataframes/105de8c7a17f4728931344dd36a28313.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6a835bffe9f71bc8570dadeb365a504e38ee0100 --- /dev/null +++ b/questions/BounceBall/dataframes/105de8c7a17f4728931344dd36a28313.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0fa692ad77578f9f5fcb6a707ec8216d7cbbcea4e1c6c5830fdc7402f98d58cd +size 52520 diff --git a/questions/BounceBall/dataframes/11e33a2095244285ad1ae8d000034bf2.parquet b/questions/BounceBall/dataframes/11e33a2095244285ad1ae8d000034bf2.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2888b422778fde6045bc9c51f47bb8269154ba9f --- /dev/null +++ b/questions/BounceBall/dataframes/11e33a2095244285ad1ae8d000034bf2.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ce427a665b2b30c39de1d4ef6c81125a597f939e1b83c34f806d8cb3a358289 +size 52545 diff --git a/questions/BounceBall/dataframes/1276239e6f2a43c58fe7ab267bd41d94.parquet b/questions/BounceBall/dataframes/1276239e6f2a43c58fe7ab267bd41d94.parquet new file mode 100644 index 0000000000000000000000000000000000000000..dc619ade60f4cc2715dc6944280ba8843a646be2 --- /dev/null +++ b/questions/BounceBall/dataframes/1276239e6f2a43c58fe7ab267bd41d94.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0380d22b2ab9fdc372e0160f10d10f6a82d6459dae9e7eeca8f2a68e4a22533e +size 52490 diff --git a/questions/BounceBall/dataframes/12dd1c92dc4649d4808421b7c2dbb124.parquet b/questions/BounceBall/dataframes/12dd1c92dc4649d4808421b7c2dbb124.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4c91ab97f7fc2a51299667ddef861e8db6ead24b --- /dev/null +++ b/questions/BounceBall/dataframes/12dd1c92dc4649d4808421b7c2dbb124.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:937eb5b7323fef7dca8d4e4afb8ebdd91ff9196b2e07383b64495fe2c3581a22 +size 52534 diff --git a/questions/BounceBall/dataframes/148cb65464985cf09acc84595e7a7ae3.parquet b/questions/BounceBall/dataframes/148cb65464985cf09acc84595e7a7ae3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9ebec476905d2e0a3ba37848f1cf38f714f280e0 --- /dev/null +++ b/questions/BounceBall/dataframes/148cb65464985cf09acc84595e7a7ae3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4556db1035ff9e46bd711908f49e130167319dfac2113d507eede9c1d7f0d1db +size 52510 diff --git a/questions/BounceBall/dataframes/1a6d28e85a7a41a292e5c95ea91ee2cb.parquet b/questions/BounceBall/dataframes/1a6d28e85a7a41a292e5c95ea91ee2cb.parquet new file mode 100644 index 0000000000000000000000000000000000000000..eecfe8c3e7fb123839fcf1d81e7e3a33d06e538d --- /dev/null +++ b/questions/BounceBall/dataframes/1a6d28e85a7a41a292e5c95ea91ee2cb.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4da48079311276c05aaf25a0937cef60e7c8a246daaef49390793e94c7537379 +size 52520 diff --git a/questions/BounceBall/dataframes/1aac2ed4a25544c894c596067aeb18ae.parquet b/questions/BounceBall/dataframes/1aac2ed4a25544c894c596067aeb18ae.parquet new file mode 100644 index 0000000000000000000000000000000000000000..82785ac3c400018d9ca297898371e25df3190d42 --- /dev/null +++ b/questions/BounceBall/dataframes/1aac2ed4a25544c894c596067aeb18ae.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0f2192c0b2502847329c9653b29a3157254d84045da38210f647efb491190d3 +size 52520 diff --git a/questions/BounceBall/dataframes/1b364086c3c740789d832cfebdd6cb1e.parquet b/questions/BounceBall/dataframes/1b364086c3c740789d832cfebdd6cb1e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9cce496ee565b43fad2496c92e1f5d3e54ec46ed --- /dev/null +++ b/questions/BounceBall/dataframes/1b364086c3c740789d832cfebdd6cb1e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9bd6e70d1f3a6a332376c0ce98d6b16667f006cd2e12b7152d17fb2e0f1622ec +size 52530 diff --git a/questions/BounceBall/dataframes/1e6654d17d454951848ea2e61337a186.parquet b/questions/BounceBall/dataframes/1e6654d17d454951848ea2e61337a186.parquet new file mode 100644 index 0000000000000000000000000000000000000000..09bef3a2bec3904400a1cbabc04d20f05ad048ca --- /dev/null +++ b/questions/BounceBall/dataframes/1e6654d17d454951848ea2e61337a186.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e68314040d7cb02f053fd1265b9bb35d007adda8aa869893e9884b464854979f +size 52520 diff --git a/questions/BounceBall/dataframes/22bd886147e847e9a0420134bbfcd2c8.parquet b/questions/BounceBall/dataframes/22bd886147e847e9a0420134bbfcd2c8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d732d1b9b3808fb54714154626c9d95c5da26811 --- /dev/null +++ b/questions/BounceBall/dataframes/22bd886147e847e9a0420134bbfcd2c8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d8958547cf1cae89c43e2d551f5c98d1d5ecc3dd4a77421dc2d08b64701e6f6 +size 52534 diff --git a/questions/BounceBall/dataframes/2310e88aa2a541b4a3a573100f28dcad.parquet b/questions/BounceBall/dataframes/2310e88aa2a541b4a3a573100f28dcad.parquet new file mode 100644 index 0000000000000000000000000000000000000000..569e2d03a1cb788e04849db5a5bb6a657103adf4 --- /dev/null +++ b/questions/BounceBall/dataframes/2310e88aa2a541b4a3a573100f28dcad.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f40bcf473a51d90f1d417413a9b6e79a134729f869ed7939f3efbfcbcad531e0 +size 52519 diff --git a/questions/BounceBall/dataframes/245d11d70caf412eb50b80db74cfb5a0.parquet b/questions/BounceBall/dataframes/245d11d70caf412eb50b80db74cfb5a0.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c86a898d1e4d210606a0b0efb717c26f2992a429 --- /dev/null +++ b/questions/BounceBall/dataframes/245d11d70caf412eb50b80db74cfb5a0.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5a0e1176bef8f4903a0a7e322a71609ae98672664d6b9e6c5135a63290d24fd +size 52595 diff --git a/questions/BounceBall/dataframes/25929fbc910e457d9143e2899419263b.parquet b/questions/BounceBall/dataframes/25929fbc910e457d9143e2899419263b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..efa77a62c07543871198f4e3f1d4715820a70814 --- /dev/null +++ b/questions/BounceBall/dataframes/25929fbc910e457d9143e2899419263b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad472792178c15b5eff7024d66e6f1f7372a1e4f3d99cf38836e7d668743bcbe +size 52520 diff --git a/questions/BounceBall/dataframes/266ff8b228ea4c2bbc98b40b4b29ae1e.parquet b/questions/BounceBall/dataframes/266ff8b228ea4c2bbc98b40b4b29ae1e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ea233a12950d01c00d6a18f0f8b449ab224dbca9 --- /dev/null +++ b/questions/BounceBall/dataframes/266ff8b228ea4c2bbc98b40b4b29ae1e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ddfa8579dd53d49cd2dedf8015e7a99019c12022244886198501cc23b1ba491 +size 52540 diff --git a/questions/BounceBall/dataframes/2736666ef08051b18037717bad48edcc.parquet b/questions/BounceBall/dataframes/2736666ef08051b18037717bad48edcc.parquet new file mode 100644 index 0000000000000000000000000000000000000000..fde2688db84fe8c48cb8cdfbe448be00fdd9b229 --- /dev/null +++ b/questions/BounceBall/dataframes/2736666ef08051b18037717bad48edcc.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1703b51031e4e2088a7ae5cab6f3fba948cfe6c135f2dc33698d2567c6aa5249 +size 52541 diff --git a/questions/BounceBall/dataframes/2888456b9ece4be3bd11a8cc7dbf349b.parquet b/questions/BounceBall/dataframes/2888456b9ece4be3bd11a8cc7dbf349b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6da7a123507bca064176497342c4ca388fc38146 --- /dev/null +++ b/questions/BounceBall/dataframes/2888456b9ece4be3bd11a8cc7dbf349b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e2ea09b44b80365c233c96aebeae55c5c0042d9fc5e40a6e964ebe139ae0a4c +size 52520 diff --git a/questions/BounceBall/dataframes/29a0305a7d8947b0983ab26ff8f5b47a.parquet b/questions/BounceBall/dataframes/29a0305a7d8947b0983ab26ff8f5b47a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..25129db9aeca1fcbef5597c3daa2718d6bbd6447 --- /dev/null +++ b/questions/BounceBall/dataframes/29a0305a7d8947b0983ab26ff8f5b47a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dee6f9445a90c0dedff272f4c73d58e44d01ab9a9ce4490153e1934158142227 +size 52520 diff --git a/questions/BounceBall/dataframes/2afe39e081a14923a1d46c052479dc3e.parquet b/questions/BounceBall/dataframes/2afe39e081a14923a1d46c052479dc3e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5eb8191ab95c943b5c645b244c81df2a1a6e2ffa --- /dev/null +++ b/questions/BounceBall/dataframes/2afe39e081a14923a1d46c052479dc3e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45d32bd0e859c1d1960fee8a8370de2a1493e0956aa2ace3f4816a02a32adc69 +size 52520 diff --git a/questions/BounceBall/dataframes/2c5b25ed588c52f5a5b4f58fc232e4bb.parquet b/questions/BounceBall/dataframes/2c5b25ed588c52f5a5b4f58fc232e4bb.parquet new file mode 100644 index 0000000000000000000000000000000000000000..584cad1865161b6aabfc99301215522fda666e9a --- /dev/null +++ b/questions/BounceBall/dataframes/2c5b25ed588c52f5a5b4f58fc232e4bb.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1624cf8d3b148b963139f6316680cfc6aedb0161384f6398fe989e52f2c855c3 +size 52534 diff --git a/questions/BounceBall/dataframes/2f760a132640578b96960b7cee8ae277.parquet b/questions/BounceBall/dataframes/2f760a132640578b96960b7cee8ae277.parquet new file mode 100644 index 0000000000000000000000000000000000000000..947cd0e5effb81c109518900f8427fe1692f27f6 --- /dev/null +++ b/questions/BounceBall/dataframes/2f760a132640578b96960b7cee8ae277.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a110584b2abeba99068a434f59d1ca8031e8ab1975d5b9785c27acf6a140cb4d +size 52510 diff --git a/questions/BounceBall/dataframes/324f1712b21d47888e27b3d794abc71b.parquet b/questions/BounceBall/dataframes/324f1712b21d47888e27b3d794abc71b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..15922c27ac164d9a27a246888f622ade5aa59d82 --- /dev/null +++ b/questions/BounceBall/dataframes/324f1712b21d47888e27b3d794abc71b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9590c9b9a33cdf4e7a6f8e9774b0233070e38212541217e241047180a37a0475 +size 52533 diff --git a/questions/BounceBall/dataframes/35428ae77661442ab6f33894965097ef.parquet b/questions/BounceBall/dataframes/35428ae77661442ab6f33894965097ef.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e625eab3438ab6d1266dfe6a902729a99d861178 --- /dev/null +++ b/questions/BounceBall/dataframes/35428ae77661442ab6f33894965097ef.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6539a549d09f844f95a619d3b75e988337a7e6f0e9e2602d0c4185629c609893 +size 52520 diff --git a/questions/BounceBall/dataframes/354dbe3c5b4b4ef29b2e2a1a0ae5c035.parquet b/questions/BounceBall/dataframes/354dbe3c5b4b4ef29b2e2a1a0ae5c035.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6da3c6417cb3f07ef4436648ee254b729eb7f11d --- /dev/null +++ b/questions/BounceBall/dataframes/354dbe3c5b4b4ef29b2e2a1a0ae5c035.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c000d4c6a5a839de516103eae0d1e2cf71c8f6736ca3b5fcc5a097265086082f +size 52545 diff --git a/questions/BounceBall/dataframes/35ce9600f2b649f39794bad12c8451f5.parquet b/questions/BounceBall/dataframes/35ce9600f2b649f39794bad12c8451f5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..dcd44df91cff6cc0c6dda097945b865c5bef3da7 --- /dev/null +++ b/questions/BounceBall/dataframes/35ce9600f2b649f39794bad12c8451f5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86ea2fc998abc8f1f9e2d81ef11d99996ba96e36dcc497f67ddfd9078a64c7f8 +size 52563 diff --git a/questions/BounceBall/dataframes/3731111840514a3a84fa9d182c7fb512.parquet b/questions/BounceBall/dataframes/3731111840514a3a84fa9d182c7fb512.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9614aa5cd97b10ba85db6f41a24917a7c686344f --- /dev/null +++ b/questions/BounceBall/dataframes/3731111840514a3a84fa9d182c7fb512.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bcc8fdef7d7e04d7a5bcaab81f58558fb2a37902d7410271dce30fb5410e963 +size 52548 diff --git a/questions/BounceBall/dataframes/3828621236495616b794ce84cf8c84ea.parquet b/questions/BounceBall/dataframes/3828621236495616b794ce84cf8c84ea.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c1d480e1b751a5f3fb98259589c11f519010769e --- /dev/null +++ b/questions/BounceBall/dataframes/3828621236495616b794ce84cf8c84ea.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fd352d722200a06955bf1d17cdb12bb9fec2bcbf7ecee2ed6f73bd6b5087b6c +size 52520 diff --git a/questions/BounceBall/dataframes/3a901bfee163599cb81f0751418e5ece.parquet b/questions/BounceBall/dataframes/3a901bfee163599cb81f0751418e5ece.parquet new file mode 100644 index 0000000000000000000000000000000000000000..33330aa15417bd198eb1b56a8ff6f09b79b660f9 --- /dev/null +++ b/questions/BounceBall/dataframes/3a901bfee163599cb81f0751418e5ece.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af4fc0d8e6c04e3c61b2baab5725fda635482d63956f85e3e293bbc862310c60 +size 52950 diff --git a/questions/BounceBall/dataframes/3b00a79d1fff4b02a8566c05ad618150.parquet b/questions/BounceBall/dataframes/3b00a79d1fff4b02a8566c05ad618150.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0ceba6c47b72a4d19ff4cc2e8fc140475ad0c56e --- /dev/null +++ b/questions/BounceBall/dataframes/3b00a79d1fff4b02a8566c05ad618150.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9534bee95a5c7c9825ee4a07298b95bf517b5740e3d65bf7da3dd3f6b125377c +size 52542 diff --git a/questions/BounceBall/dataframes/3f14e6ab62885cecad3876e7b3caf6dd.parquet b/questions/BounceBall/dataframes/3f14e6ab62885cecad3876e7b3caf6dd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..611501fd842e9fe0b712189a1867f0637e712fb1 --- /dev/null +++ b/questions/BounceBall/dataframes/3f14e6ab62885cecad3876e7b3caf6dd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25a7da0cd333cefd0df58f706f4a634cf19b71c2637378b4b37ac9305b92c2b4 +size 52510 diff --git a/questions/BounceBall/dataframes/3fd9c7f94e355f65bada44cc585647d6.parquet b/questions/BounceBall/dataframes/3fd9c7f94e355f65bada44cc585647d6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f2f5da7adf859c31adb5784d3c9895db10234ded --- /dev/null +++ b/questions/BounceBall/dataframes/3fd9c7f94e355f65bada44cc585647d6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e8d7a75b7be8bb08cb0f0a59aa5cef5e2b99c320f3a78a6a9a6fb57d5010151 +size 52534 diff --git a/questions/BounceBall/dataframes/3ff3d291e19c586a958adb1f94864c17.parquet b/questions/BounceBall/dataframes/3ff3d291e19c586a958adb1f94864c17.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8afac69db8261da7fb8617a463d06b21dae814e3 --- /dev/null +++ b/questions/BounceBall/dataframes/3ff3d291e19c586a958adb1f94864c17.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a5a1973cbc6b032dc356e63ed39e464a108dc4391e4c88cfb5e356549c20e75 +size 52530 diff --git a/questions/BounceBall/dataframes/42a04eb48f2359898f15b1581dab6727.parquet b/questions/BounceBall/dataframes/42a04eb48f2359898f15b1581dab6727.parquet new file mode 100644 index 0000000000000000000000000000000000000000..358b7b2b98a50b53394cdcf1b6ea50101fa73e8e --- /dev/null +++ b/questions/BounceBall/dataframes/42a04eb48f2359898f15b1581dab6727.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a7ae61844569e79d325ae0dc14bbe9986cff2b6286662fa54b89c58cae22cb33 +size 52510 diff --git a/questions/BounceBall/dataframes/4483b5b67b3f4d9abae899f3bea83fa0.parquet b/questions/BounceBall/dataframes/4483b5b67b3f4d9abae899f3bea83fa0.parquet new file mode 100644 index 0000000000000000000000000000000000000000..298f32b67d8f11ea4b23aad28e529f62524b56b2 --- /dev/null +++ b/questions/BounceBall/dataframes/4483b5b67b3f4d9abae899f3bea83fa0.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8267de5b950d0e71f0170e973173924590cf7581d263280bac7a02b54713f5ab +size 43843 diff --git a/questions/BounceBall/dataframes/45aa02ea71c0510aae5f198aef30fef5.parquet b/questions/BounceBall/dataframes/45aa02ea71c0510aae5f198aef30fef5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f2c8538de12c830d89874b5544ec10e5e5a35e16 --- /dev/null +++ b/questions/BounceBall/dataframes/45aa02ea71c0510aae5f198aef30fef5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:437dd67d25143bb8a0505c2a7103ecd3cec0ccb4b9630330702d6fed5939f5a7 +size 52510 diff --git a/questions/BounceBall/dataframes/461103bd3ac255b2a1e6bd5a4122cbc4.parquet b/questions/BounceBall/dataframes/461103bd3ac255b2a1e6bd5a4122cbc4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7c426ffe8ccd9d7a3364d21cf75c33f7854c20c2 --- /dev/null +++ b/questions/BounceBall/dataframes/461103bd3ac255b2a1e6bd5a4122cbc4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:81e81c1433fae98c9836ba765c573fb6f2323cf308ce7e96be77eb38ab0b2efe +size 52649 diff --git a/questions/BounceBall/dataframes/4bdd20ee31ce447eb8b038067bdd66d7.parquet b/questions/BounceBall/dataframes/4bdd20ee31ce447eb8b038067bdd66d7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ca1a5dad29b9d90b8d23381e3f4d2306aa3d9f98 --- /dev/null +++ b/questions/BounceBall/dataframes/4bdd20ee31ce447eb8b038067bdd66d7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:53c59ab7120858fdd180f39429932a233aff43737f9f1a5f2932d96aed3016ef +size 52530 diff --git a/questions/BounceBall/dataframes/4d8fb279215248f8acbfad4f5e395b2e.parquet b/questions/BounceBall/dataframes/4d8fb279215248f8acbfad4f5e395b2e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bd4043e0d9485a90e63166c74e26383df2df747b --- /dev/null +++ b/questions/BounceBall/dataframes/4d8fb279215248f8acbfad4f5e395b2e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85a9bdb72ac9fdceb4b0a9cce5ec98b042ef3b474562a64d417cd21c543bd381 +size 52520 diff --git a/questions/BounceBall/dataframes/4e53c84aa5024bf4ae26bb23d7a4d07e.parquet b/questions/BounceBall/dataframes/4e53c84aa5024bf4ae26bb23d7a4d07e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..726e1d0bf80cc09caebdc9f8342ec6c5045c00c7 --- /dev/null +++ b/questions/BounceBall/dataframes/4e53c84aa5024bf4ae26bb23d7a4d07e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c67fa2d165474eb07ee2c7c1bef3c618bb1639bc77b5fe29f62a771510e7b911 +size 52520 diff --git a/questions/BounceBall/dataframes/4f26efbd7d274f7fa043588a6d4e7204.parquet b/questions/BounceBall/dataframes/4f26efbd7d274f7fa043588a6d4e7204.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c5a14b543f73c46e6da71909599f525c7ab00d50 --- /dev/null +++ b/questions/BounceBall/dataframes/4f26efbd7d274f7fa043588a6d4e7204.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7637c6645d93b46eb4e2172b614550adcca2b01a5dd4b8c611e2f4a9862d9d2 +size 52584 diff --git a/questions/BounceBall/dataframes/4f60288e7a90582a841638ff2aaca5cc.parquet b/questions/BounceBall/dataframes/4f60288e7a90582a841638ff2aaca5cc.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7c8f77c5906ad117d7f3b7e809020b9d441a4ee6 --- /dev/null +++ b/questions/BounceBall/dataframes/4f60288e7a90582a841638ff2aaca5cc.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:614ce242825ed26232ba0b86c92190319ae3063ecb358137301955a2cc5f867b +size 52534 diff --git a/questions/BounceBall/dataframes/55cd6c88ee704897a0318d3680135fb7.parquet b/questions/BounceBall/dataframes/55cd6c88ee704897a0318d3680135fb7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..43b8248a36417e1b0343eab3d0995f3e3abfd8dc --- /dev/null +++ b/questions/BounceBall/dataframes/55cd6c88ee704897a0318d3680135fb7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f1155805c823501be77b256f73f60f35ad4446a7cce5f77ff4f9e900551a514 +size 52541 diff --git a/questions/BounceBall/dataframes/5838d4e3197541219e46a399693b4d50.parquet b/questions/BounceBall/dataframes/5838d4e3197541219e46a399693b4d50.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d88a2a7fa32f00a13ae4e788c1bc1652006a61c3 --- /dev/null +++ b/questions/BounceBall/dataframes/5838d4e3197541219e46a399693b4d50.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:debd6785b07eb77a20fbc0d88573a776398654c973e54f8a6aabd088fd01c453 +size 52538 diff --git a/questions/BounceBall/dataframes/5990d64f12064184ac3587e5004bac94.parquet b/questions/BounceBall/dataframes/5990d64f12064184ac3587e5004bac94.parquet new file mode 100644 index 0000000000000000000000000000000000000000..427e8d49196c21e47d546a5429563a8624d5b220 --- /dev/null +++ b/questions/BounceBall/dataframes/5990d64f12064184ac3587e5004bac94.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:757e153e9b4ea90929c419f1ec2a5a0495c83b6a95a37cc510d65e34cfd0ce11 +size 52510 diff --git a/questions/BounceBall/dataframes/5af136001c8e55ed96639fb169b820df.parquet b/questions/BounceBall/dataframes/5af136001c8e55ed96639fb169b820df.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ca089905768a98a80aba8882019a8d3e76ea144f --- /dev/null +++ b/questions/BounceBall/dataframes/5af136001c8e55ed96639fb169b820df.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4a877cf489df38c4595a538df7f36db4f7b5bdcb035142eb88d876a37064ba2 +size 52545 diff --git a/questions/BounceBall/dataframes/5ed1c131e780539b84a3e3c3e76d73d4.parquet b/questions/BounceBall/dataframes/5ed1c131e780539b84a3e3c3e76d73d4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e42f0656ef33f050a71453dcca36688044e5f9a2 --- /dev/null +++ b/questions/BounceBall/dataframes/5ed1c131e780539b84a3e3c3e76d73d4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91ecd2bf81b3303eb79662d787e32f62507e281b2aa7979f94392af55e0efeed +size 52506 diff --git a/questions/BounceBall/dataframes/5f6f4011eba24b46a09cac472f54fff4.parquet b/questions/BounceBall/dataframes/5f6f4011eba24b46a09cac472f54fff4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4002c72d2d14ff04d6847a5ad5d508a6f774270d --- /dev/null +++ b/questions/BounceBall/dataframes/5f6f4011eba24b46a09cac472f54fff4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d78477f71aa7dbb0a7ce5f60d85f7821ddb6dbbd5ad5535aaff49267445af991 +size 52510 diff --git a/questions/BounceBall/dataframes/6407da2f934a51c5951cb7161ace6674.parquet b/questions/BounceBall/dataframes/6407da2f934a51c5951cb7161ace6674.parquet new file mode 100644 index 0000000000000000000000000000000000000000..09301c92cc1118b82204d874793b55e29e3d593b --- /dev/null +++ b/questions/BounceBall/dataframes/6407da2f934a51c5951cb7161ace6674.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5c1c6adf56e6fd7f1cb9eeb72523643d97231715e534afb9692c682a45d1599 +size 52520 diff --git a/questions/BounceBall/dataframes/664fa7515eea439692b7f460afbf1631.parquet b/questions/BounceBall/dataframes/664fa7515eea439692b7f460afbf1631.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c81409a10b7da69048c3d0af5e0e2764167b3601 --- /dev/null +++ b/questions/BounceBall/dataframes/664fa7515eea439692b7f460afbf1631.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:edcd49d788a54b5569b71821aadc2723ddda5c1cd5d70ab66f55f1a9747ef126 +size 52520 diff --git a/questions/BounceBall/dataframes/6783ef519fdc4d7bb71f66db00c264e7.parquet b/questions/BounceBall/dataframes/6783ef519fdc4d7bb71f66db00c264e7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5c2dad06fe78954c969091288a59d751f4ce0814 --- /dev/null +++ b/questions/BounceBall/dataframes/6783ef519fdc4d7bb71f66db00c264e7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f97a2dfbf31cf11fb450c68d426155c75a3ea15b0660266dccb9fa3528a7c49 +size 52563 diff --git a/questions/BounceBall/dataframes/68c17889e0f146f7a0772ecd1c0e63cf.parquet b/questions/BounceBall/dataframes/68c17889e0f146f7a0772ecd1c0e63cf.parquet new file mode 100644 index 0000000000000000000000000000000000000000..19f0bfdd1eafec26dd95eb4bf6a19a087cdbf65b --- /dev/null +++ b/questions/BounceBall/dataframes/68c17889e0f146f7a0772ecd1c0e63cf.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3104123d1d2e02583133c473418b7e9c32ce35ff51bb01b612da8802061bc797 +size 52533 diff --git a/questions/BounceBall/dataframes/6be5dc20d4365c68ae8ef3bae8a17615.parquet b/questions/BounceBall/dataframes/6be5dc20d4365c68ae8ef3bae8a17615.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bdcaf26912e927cce34c6676278726ceaa8d9090 --- /dev/null +++ b/questions/BounceBall/dataframes/6be5dc20d4365c68ae8ef3bae8a17615.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3bd4e70128a8c7155a9414af43f31fdd81d7b61634a22715e1f865dee807079e +size 52520 diff --git a/questions/BounceBall/dataframes/6c4ff7108da34c2d94ba2bb300dd1ee7.parquet b/questions/BounceBall/dataframes/6c4ff7108da34c2d94ba2bb300dd1ee7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..48e46f18a804039c5eb308aab48b6dbe5a6036b2 --- /dev/null +++ b/questions/BounceBall/dataframes/6c4ff7108da34c2d94ba2bb300dd1ee7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:00d8501430c8deace691bf77261e32f845d38671305aacf403c1b173763d6bc1 +size 52564 diff --git a/questions/BounceBall/dataframes/6ca1f93c5f9d59669a10db584518a5da.parquet b/questions/BounceBall/dataframes/6ca1f93c5f9d59669a10db584518a5da.parquet new file mode 100644 index 0000000000000000000000000000000000000000..31b1391a9d4306c92b90be0f56e9ba327fd9d120 --- /dev/null +++ b/questions/BounceBall/dataframes/6ca1f93c5f9d59669a10db584518a5da.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:776b5c21ef0e6830cba6af021fdd4befe0a29e7383723e65f9d0f0efd43436e6 +size 51291 diff --git a/questions/BounceBall/dataframes/7379873992335258a1c5571e0f7902db.parquet b/questions/BounceBall/dataframes/7379873992335258a1c5571e0f7902db.parquet new file mode 100644 index 0000000000000000000000000000000000000000..61d1f17246f64c9c94cc47be1e474809f3a9603b --- /dev/null +++ b/questions/BounceBall/dataframes/7379873992335258a1c5571e0f7902db.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6717de6bae5028afb8df544b7d793918a935f77561b383bbf350e78619204bf +size 52541 diff --git a/questions/BounceBall/dataframes/73e57408da7a53bba190eae95592a58b.parquet b/questions/BounceBall/dataframes/73e57408da7a53bba190eae95592a58b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..76a4deef3670d1457d1cc0652977daae8f01c156 --- /dev/null +++ b/questions/BounceBall/dataframes/73e57408da7a53bba190eae95592a58b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff947daf16f0befd51b80ec84a31c7eef061164016caf8172cd5766f8f746a25 +size 52510 diff --git a/questions/BounceBall/dataframes/74b7a23252a94ac4aa7aa071d7b17aa5.parquet b/questions/BounceBall/dataframes/74b7a23252a94ac4aa7aa071d7b17aa5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ccd9814bc57eeba9ee9a2a5b26ba905e335a98c3 --- /dev/null +++ b/questions/BounceBall/dataframes/74b7a23252a94ac4aa7aa071d7b17aa5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3f7c4bb31682756517865df25822d616366cac68faf93271b0b386018b8fdda1 +size 52542 diff --git a/questions/BounceBall/dataframes/779261e592034cf882881c2f70a20728.parquet b/questions/BounceBall/dataframes/779261e592034cf882881c2f70a20728.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2ea91f8f8b14d35ee257f6434342a4f3eb6a0e85 --- /dev/null +++ b/questions/BounceBall/dataframes/779261e592034cf882881c2f70a20728.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca75fcfc94d5e27fb4e777c1b57c2df9530b260bba97f7daee3806f38234adca +size 52533 diff --git a/questions/BounceBall/dataframes/78553d66439e5f678fe0b5574efa7384.parquet b/questions/BounceBall/dataframes/78553d66439e5f678fe0b5574efa7384.parquet new file mode 100644 index 0000000000000000000000000000000000000000..fd334396e1b39bf76f6c21f983ba293a0b072745 --- /dev/null +++ b/questions/BounceBall/dataframes/78553d66439e5f678fe0b5574efa7384.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87308787a87b4b14414ae07ce539141046461bb00cb351cbe355bf428acf10a5 +size 52545 diff --git a/questions/BounceBall/dataframes/7b2b2d6cce1c444d8fc6bc0be9ac42ee.parquet b/questions/BounceBall/dataframes/7b2b2d6cce1c444d8fc6bc0be9ac42ee.parquet new file mode 100644 index 0000000000000000000000000000000000000000..524a159157583cab6251cb8ee312e1042e2385d5 --- /dev/null +++ b/questions/BounceBall/dataframes/7b2b2d6cce1c444d8fc6bc0be9ac42ee.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:351eea0b6c84d952a037dbc5df27cad386d29c0fbccecdb6b6b546d1d4ffe17b +size 52545 diff --git a/questions/BounceBall/dataframes/7b377fd33ec55476b4dd35ce0f6adf5b.parquet b/questions/BounceBall/dataframes/7b377fd33ec55476b4dd35ce0f6adf5b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ff1bbbeee90f25c77966f0b2da73597ac2caf9ee --- /dev/null +++ b/questions/BounceBall/dataframes/7b377fd33ec55476b4dd35ce0f6adf5b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:36f6a69a449ab6d54f452c32df1938675462712d4bcfa067bf07bf1d3d53c6bb +size 52585 diff --git a/questions/BounceBall/dataframes/7dacaf46a9ec45cf9c7e1aabad31e83f.parquet b/questions/BounceBall/dataframes/7dacaf46a9ec45cf9c7e1aabad31e83f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ee1f9bcd53739d46a6312d3ba4dc5b0304964af2 --- /dev/null +++ b/questions/BounceBall/dataframes/7dacaf46a9ec45cf9c7e1aabad31e83f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29946f25f4cbc5018029f1f46a8e47c16417427d367616549fbad1d96fa4d980 +size 52530 diff --git a/questions/BounceBall/dataframes/81064a33c383447293b0d82c15f891bb.parquet b/questions/BounceBall/dataframes/81064a33c383447293b0d82c15f891bb.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6e3c1a783a2822f7a4ddc8287cdce908dec8ae70 --- /dev/null +++ b/questions/BounceBall/dataframes/81064a33c383447293b0d82c15f891bb.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e487f848d1cacd7e99e315d5530f364e1f0b409e9d8057abe7d027e4b4194c23 +size 52563 diff --git a/questions/BounceBall/dataframes/8399e56763da5bc59ef9f9c4b652eeee.parquet b/questions/BounceBall/dataframes/8399e56763da5bc59ef9f9c4b652eeee.parquet new file mode 100644 index 0000000000000000000000000000000000000000..78ba7b3347244ce92b2de2308d3cfd78eea907ae --- /dev/null +++ b/questions/BounceBall/dataframes/8399e56763da5bc59ef9f9c4b652eeee.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9ed9faa653afe25cec915915118c621f297372c37e2d3ed7a8883e4a2ff19194 +size 52744 diff --git a/questions/BounceBall/dataframes/84bbd6c06ddb4b54a3f12c882c2eccc4.parquet b/questions/BounceBall/dataframes/84bbd6c06ddb4b54a3f12c882c2eccc4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..25bd5da12c7c3195c70ca3da9d5358419e3aea5b --- /dev/null +++ b/questions/BounceBall/dataframes/84bbd6c06ddb4b54a3f12c882c2eccc4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ad4f1b5338015aa0b8e39f21e790694cc5dc71931722015193d61567f96171b +size 52534 diff --git a/questions/BounceBall/dataframes/8549e70b7f975ae68b059707d6673904.parquet b/questions/BounceBall/dataframes/8549e70b7f975ae68b059707d6673904.parquet new file mode 100644 index 0000000000000000000000000000000000000000..29895f618ba78d3e9e542bf6f3f9ff52047b73c0 --- /dev/null +++ b/questions/BounceBall/dataframes/8549e70b7f975ae68b059707d6673904.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:81c042a91fdca1da72c466314d1b77941fb9a1bafb2cdb4604e653c473711584 +size 52541 diff --git a/questions/BounceBall/dataframes/855c0b4b25c947ddada661a7b2138f42.parquet b/questions/BounceBall/dataframes/855c0b4b25c947ddada661a7b2138f42.parquet new file mode 100644 index 0000000000000000000000000000000000000000..87db4010d6860f675f293a66a4b397f22eeca559 --- /dev/null +++ b/questions/BounceBall/dataframes/855c0b4b25c947ddada661a7b2138f42.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7212f9e4e75f2d4b8012934edb470bdaa684c2c673021ccb91d353389b28cd3e +size 52520 diff --git a/questions/BounceBall/dataframes/866e78a3aee946dfaab06893eac3d553.parquet b/questions/BounceBall/dataframes/866e78a3aee946dfaab06893eac3d553.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c9338bd53caac9b5d5858e678e33ec1acfa5575e --- /dev/null +++ b/questions/BounceBall/dataframes/866e78a3aee946dfaab06893eac3d553.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02145050b10800098559c2d0dc9d1b491de99796025cd1217a58ceb8c1ec208b +size 52534 diff --git a/questions/BounceBall/dataframes/8f80071112174821a3595067c2bf1537.parquet b/questions/BounceBall/dataframes/8f80071112174821a3595067c2bf1537.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cdc18c2c47dd3515c637f6e09554e4e4165c7bcd --- /dev/null +++ b/questions/BounceBall/dataframes/8f80071112174821a3595067c2bf1537.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4136e630f94c11c960c08c16c377d1d5a2af644556be9fbdef33b0f2041125c8 +size 52520 diff --git a/questions/BounceBall/dataframes/90872800e402540eb77c39d26a123b23.parquet b/questions/BounceBall/dataframes/90872800e402540eb77c39d26a123b23.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d2f7e24d182b4b45a643231d7f7f2bd7a6767e38 --- /dev/null +++ b/questions/BounceBall/dataframes/90872800e402540eb77c39d26a123b23.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9afd08467f9dc7f45f3d3e04d2cec12f8a25ac93b5c1cc92f95b2f16384a2376 +size 52520 diff --git a/questions/BounceBall/dataframes/934a4597164b46a7a2728e1ebf583b32.parquet b/questions/BounceBall/dataframes/934a4597164b46a7a2728e1ebf583b32.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5aa152f7644e9a4d04dc00f1b53c5687e9ae18ed --- /dev/null +++ b/questions/BounceBall/dataframes/934a4597164b46a7a2728e1ebf583b32.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:743519e6f59ed5026664fb6e6d7d61119bfd34d84aa017e81b97141622419374 +size 52553 diff --git a/questions/BounceBall/dataframes/9584da5231b4512188b5089c9a0dbc05.parquet b/questions/BounceBall/dataframes/9584da5231b4512188b5089c9a0dbc05.parquet new file mode 100644 index 0000000000000000000000000000000000000000..35116318a9a10e350a6aa16d8d224861f9843773 --- /dev/null +++ b/questions/BounceBall/dataframes/9584da5231b4512188b5089c9a0dbc05.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b73ca25b4c551803924eeac64306178493e18255bfe5d85371f1d1dd933ab21 +size 52587 diff --git a/questions/BounceBall/dataframes/95ed9fef81a25479881312e047965573.parquet b/questions/BounceBall/dataframes/95ed9fef81a25479881312e047965573.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5e3463024ea2b8b5e0e024e24c6d67e3d3bcfcf7 --- /dev/null +++ b/questions/BounceBall/dataframes/95ed9fef81a25479881312e047965573.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:791a94a93a312acd50387a8bddd57f2145d7fc814f0d9acc0373a825d55f7208 +size 52520 diff --git a/questions/BounceBall/dataframes/9698d8207595571e9187650324239bb0.parquet b/questions/BounceBall/dataframes/9698d8207595571e9187650324239bb0.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d5d86f6e81807d7644f4e107299f16f970afe09e --- /dev/null +++ b/questions/BounceBall/dataframes/9698d8207595571e9187650324239bb0.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28918713e01e2d87ba10ccb51e8b1dc43014d27b70a97a456a5a6a38e62f274b +size 52533 diff --git a/questions/BounceBall/dataframes/96f1621029835fd49654f3249386c245.parquet b/questions/BounceBall/dataframes/96f1621029835fd49654f3249386c245.parquet new file mode 100644 index 0000000000000000000000000000000000000000..91138abdb5c0cf63d418dc56832e2048a1bfeae1 --- /dev/null +++ b/questions/BounceBall/dataframes/96f1621029835fd49654f3249386c245.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e1d6282438a992a5ce9d4d8776a321a231ac82c048f27b6fcb3e49333acab17 +size 52587 diff --git a/questions/BounceBall/dataframes/9820ec472b7e4371a4d40be054b4c6ae.parquet b/questions/BounceBall/dataframes/9820ec472b7e4371a4d40be054b4c6ae.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d4dd00a11ccf52558c83f6076a8fdba72df113ee --- /dev/null +++ b/questions/BounceBall/dataframes/9820ec472b7e4371a4d40be054b4c6ae.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3385cf71e09ed2772a81be35df4caa26e71a07d20f5cbf4730a0a84b593d2d35 +size 52534 diff --git a/questions/BounceBall/dataframes/98ce3759ae534c9f8ee938e151b01032.parquet b/questions/BounceBall/dataframes/98ce3759ae534c9f8ee938e151b01032.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b4b6861a7a90d2df7025e3f061691f1cef6a22d7 --- /dev/null +++ b/questions/BounceBall/dataframes/98ce3759ae534c9f8ee938e151b01032.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f6af053969e7edb9cc67dbb4265d42aa9059dddfac14fd0ae4936414175776f +size 52530 diff --git a/questions/BounceBall/dataframes/9a03fccac5e1443fbea77e7243b73b84.parquet b/questions/BounceBall/dataframes/9a03fccac5e1443fbea77e7243b73b84.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4740ea342216070c70241914007d673d80110942 --- /dev/null +++ b/questions/BounceBall/dataframes/9a03fccac5e1443fbea77e7243b73b84.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:419269f3e4f1a1e73276aec93417f429cea9e8da4082a717e78b020aee0e72cb +size 52520 diff --git a/questions/BounceBall/dataframes/a0399f67cb5f4165a4dc00fa72104bb6.parquet b/questions/BounceBall/dataframes/a0399f67cb5f4165a4dc00fa72104bb6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bc4dd7c239b5aaf2cab9805ca27966c5571ed498 --- /dev/null +++ b/questions/BounceBall/dataframes/a0399f67cb5f4165a4dc00fa72104bb6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5970f215e17c4fc23b21205d27416be346a430c4edac0a1e8367831065f269e4 +size 52533 diff --git a/questions/BounceBall/dataframes/a04e965c08014d239d8b8a393e1a3cdd.parquet b/questions/BounceBall/dataframes/a04e965c08014d239d8b8a393e1a3cdd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0645b0ecb723e27f95370567f218511768602ff6 --- /dev/null +++ b/questions/BounceBall/dataframes/a04e965c08014d239d8b8a393e1a3cdd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e899825fe4ca03eda9819afe324bbbc073c1d085d3d31599fc57152fbc4a5d5 +size 52556 diff --git a/questions/BounceBall/dataframes/a3d4d1a02f6f4d74a743b6f0212ec691.parquet b/questions/BounceBall/dataframes/a3d4d1a02f6f4d74a743b6f0212ec691.parquet new file mode 100644 index 0000000000000000000000000000000000000000..529e0238bd924d37243b2137f21e9ab335f87b8e --- /dev/null +++ b/questions/BounceBall/dataframes/a3d4d1a02f6f4d74a743b6f0212ec691.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb835a362e5ac4ac69e4845ae811cf71ec3c1cf9f103425119974dc93e00a041 +size 52534 diff --git a/questions/BounceBall/dataframes/a71f474e41464187af46ccbd7222cae8.parquet b/questions/BounceBall/dataframes/a71f474e41464187af46ccbd7222cae8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..638d8e27c044a083d63dfde542a08e9aa69865fb --- /dev/null +++ b/questions/BounceBall/dataframes/a71f474e41464187af46ccbd7222cae8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55c71d40ae8a6a34df2d63a6b9819e6df38db738a60c50fd3856db56e5bf8723 +size 52541 diff --git a/questions/BounceBall/dataframes/acd65eb4f78a5539a10bb7f268c00b8b.parquet b/questions/BounceBall/dataframes/acd65eb4f78a5539a10bb7f268c00b8b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c4aa1728e972d4fde64931d3248aa27f5a5183fb --- /dev/null +++ b/questions/BounceBall/dataframes/acd65eb4f78a5539a10bb7f268c00b8b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2dc40a90b65b5afb9973d3ba1ff504facad4f974762f38ccd15382317af01e3c +size 52520 diff --git a/questions/BounceBall/dataframes/b248181b23624fdabde51fb023f87d70.parquet b/questions/BounceBall/dataframes/b248181b23624fdabde51fb023f87d70.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d1de22d0ea0b554f627a07c2c2d2c4826747ea05 --- /dev/null +++ b/questions/BounceBall/dataframes/b248181b23624fdabde51fb023f87d70.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e719870576aeaa2dc0aea08112c18d695163623a860c06f12a6b49e1770130a +size 52534 diff --git a/questions/BounceBall/dataframes/b8b64d9c72a35d9486886e9b726261cf.parquet b/questions/BounceBall/dataframes/b8b64d9c72a35d9486886e9b726261cf.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a252eac42b79fe8ef696efaca0f4a3f8dde7c7eb --- /dev/null +++ b/questions/BounceBall/dataframes/b8b64d9c72a35d9486886e9b726261cf.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59ed89ba57769915180762475edd0ce634288548915482883c9c88601548e1fe +size 52520 diff --git a/questions/BounceBall/dataframes/ba1a51d50f0b5031b0f10f2b48f8f5b5.parquet b/questions/BounceBall/dataframes/ba1a51d50f0b5031b0f10f2b48f8f5b5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..177a8204b5e3d941ca0425db85683695b5bf0d97 --- /dev/null +++ b/questions/BounceBall/dataframes/ba1a51d50f0b5031b0f10f2b48f8f5b5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39b8355cb6c1718359c5fb5cd54266a6f5a2de4dc496c46f0b3cd851485b1e60 +size 52510 diff --git a/questions/BounceBall/dataframes/bb81df685f9a58e59ecb171a9cceb33e.parquet b/questions/BounceBall/dataframes/bb81df685f9a58e59ecb171a9cceb33e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7b83e3fb6125679f15676742528b5a57eedd7ddb --- /dev/null +++ b/questions/BounceBall/dataframes/bb81df685f9a58e59ecb171a9cceb33e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a676739bdfbc07f94ed53f0df46d84668f48d76994c8f1fcbb5c7cd1cc854c98 +size 52520 diff --git a/questions/BounceBall/dataframes/bbe78dcf65cf4990b574beb1df877cfc.parquet b/questions/BounceBall/dataframes/bbe78dcf65cf4990b574beb1df877cfc.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3ede969b53a6cbad63063a3996e6f4d119735c66 --- /dev/null +++ b/questions/BounceBall/dataframes/bbe78dcf65cf4990b574beb1df877cfc.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c6be058237e8bda503add59db53bf6b2d6bdc1950c17c00862b067265e2fa7d +size 52534 diff --git a/questions/BounceBall/dataframes/be12a2ede7744ecdae83f9e0d1f2a47f.parquet b/questions/BounceBall/dataframes/be12a2ede7744ecdae83f9e0d1f2a47f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c739d70f8ff937387f4c5086740428bbdbaf213f --- /dev/null +++ b/questions/BounceBall/dataframes/be12a2ede7744ecdae83f9e0d1f2a47f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb8084c04ed785fdb73990e126834a4332e2b2597e16d7c8fbcd4c1893f4affc +size 52510 diff --git a/questions/BounceBall/dataframes/be961034c82352e09e4c5952be16637a.parquet b/questions/BounceBall/dataframes/be961034c82352e09e4c5952be16637a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cc7915346743c45d47bb60004b2481691abbeea5 --- /dev/null +++ b/questions/BounceBall/dataframes/be961034c82352e09e4c5952be16637a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f93f5dc94fcacd1ef20efc7b27733e6a85e43e10d2709ad8e380d744ed417d4a +size 52534 diff --git a/questions/BounceBall/dataframes/c131520e5f80470ebb16965a0c6ba8f4.parquet b/questions/BounceBall/dataframes/c131520e5f80470ebb16965a0c6ba8f4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..138ecf915666cd29a83ee586db15ebbbf05eb246 --- /dev/null +++ b/questions/BounceBall/dataframes/c131520e5f80470ebb16965a0c6ba8f4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78833f360a867628aef0c31b371dfe478a414176177062f97a9687369e57af52 +size 52534 diff --git a/questions/BounceBall/dataframes/c32fcada64fe5036a1a843f7f423f6b4.parquet b/questions/BounceBall/dataframes/c32fcada64fe5036a1a843f7f423f6b4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8f0493ed5b718fdaaa59bc3e4dbe52b67771f3ee --- /dev/null +++ b/questions/BounceBall/dataframes/c32fcada64fe5036a1a843f7f423f6b4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a99fb5464a0ce26a3960bd939e0aef4bb7bbadacb2319c78a20e0ca39503b2f +size 52510 diff --git a/questions/BounceBall/dataframes/c3b3ddcba6ff5526bd21d1d06ee66506.parquet b/questions/BounceBall/dataframes/c3b3ddcba6ff5526bd21d1d06ee66506.parquet new file mode 100644 index 0000000000000000000000000000000000000000..98510343301537f93e8f89055bcd66f659d3e304 --- /dev/null +++ b/questions/BounceBall/dataframes/c3b3ddcba6ff5526bd21d1d06ee66506.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a295cd65c5e0868846cc85413272d52b3e997627570ec68e5f939cac2e9aeb14 +size 52545 diff --git a/questions/BounceBall/dataframes/c5b764da8f9442ebbce3b0859ed481a1.parquet b/questions/BounceBall/dataframes/c5b764da8f9442ebbce3b0859ed481a1.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2622eb26cdd2c6083b7694a63d667a72680ee97c --- /dev/null +++ b/questions/BounceBall/dataframes/c5b764da8f9442ebbce3b0859ed481a1.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef416f340608000a11adf946c3177ebae1db34a6baa83def4bb1851323d5b1f4 +size 52520 diff --git a/questions/BounceBall/dataframes/c67fd00576b04ac2bb3baf818c034ae4.parquet b/questions/BounceBall/dataframes/c67fd00576b04ac2bb3baf818c034ae4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..017b0824d41bf979cf8e152670514b20a2750102 --- /dev/null +++ b/questions/BounceBall/dataframes/c67fd00576b04ac2bb3baf818c034ae4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a81a941720a4df5385661bf03717f44bfaa361e14f205d38ebabe73f59b6c2d +size 52534 diff --git a/questions/BounceBall/dataframes/c7647d63ffa142a181f3bc6d78795879.parquet b/questions/BounceBall/dataframes/c7647d63ffa142a181f3bc6d78795879.parquet new file mode 100644 index 0000000000000000000000000000000000000000..77b54f6f4a3a0de4fcb71f7c2d108a2b8c84f8a0 --- /dev/null +++ b/questions/BounceBall/dataframes/c7647d63ffa142a181f3bc6d78795879.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d4329ec92deee5d020f2093e3444e649364a218d3f9c56c288d7f7fab965019 +size 52552 diff --git a/questions/BounceBall/dataframes/ca8e6694b045476e8d251f55f5445ab3.parquet b/questions/BounceBall/dataframes/ca8e6694b045476e8d251f55f5445ab3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9cc227856f06324ec524e092d87920b60b23ca98 --- /dev/null +++ b/questions/BounceBall/dataframes/ca8e6694b045476e8d251f55f5445ab3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03cadc4dc878be0379a70cd413a33d6c726a579ed2f7e3c744e098ba8be2eeea +size 52533 diff --git a/questions/BounceBall/dataframes/cc080755498055de93cebb72440a86c6.parquet b/questions/BounceBall/dataframes/cc080755498055de93cebb72440a86c6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..58621f95cf3467bf82bae4f53b7331ffa8ab11a6 --- /dev/null +++ b/questions/BounceBall/dataframes/cc080755498055de93cebb72440a86c6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc7f542ee092b7a962755a1dd43f88eebe13fdffe7af442159ec722cbec7c9ab +size 52510 diff --git a/questions/BounceBall/dataframes/cc34c415d1465097ae993175b76900b8.parquet b/questions/BounceBall/dataframes/cc34c415d1465097ae993175b76900b8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b7ee08d1fc75630cdf42457010fbb1a45b549ee1 --- /dev/null +++ b/questions/BounceBall/dataframes/cc34c415d1465097ae993175b76900b8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34368edd6d6eacaf8b0f35c7399016821abd9f41b0a74bb30472199508922d9e +size 52562 diff --git a/questions/BounceBall/dataframes/d107128850bd502784cdb511d7efa744.parquet b/questions/BounceBall/dataframes/d107128850bd502784cdb511d7efa744.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1f04c377ecf56f5548961641a5069c0c71edb48a --- /dev/null +++ b/questions/BounceBall/dataframes/d107128850bd502784cdb511d7efa744.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16eaf3f11c50dbed55d42efe9ff450a2dccb6dfc7ea07a01d2e6586f75a2048d +size 52509 diff --git a/questions/BounceBall/dataframes/d217e2d448755e4bb51130635edf22b9.parquet b/questions/BounceBall/dataframes/d217e2d448755e4bb51130635edf22b9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ac31a6dc0aa0a4bde328304b80d2906a1e874f91 --- /dev/null +++ b/questions/BounceBall/dataframes/d217e2d448755e4bb51130635edf22b9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d42c199415d880b642e0a74846ba9bb5ab0f215e2b860b41d78515d790a2fab +size 52552 diff --git a/questions/BounceBall/dataframes/d31bac7398034409ad2a8d05cf5ea3f8.parquet b/questions/BounceBall/dataframes/d31bac7398034409ad2a8d05cf5ea3f8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2067ffb542665fd4649da2064ea7a482332ed5b9 --- /dev/null +++ b/questions/BounceBall/dataframes/d31bac7398034409ad2a8d05cf5ea3f8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:230898bef3b7a9f163086a2ddc79b8f7dd6e85516a8899f394ddbc98e48b4675 +size 52516 diff --git a/questions/BounceBall/dataframes/d67bbb7d87d745ad8b5f8871e4ae9a10.parquet b/questions/BounceBall/dataframes/d67bbb7d87d745ad8b5f8871e4ae9a10.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7f2f82fb91dfe6153cda2b81bc58247fd8216f7c --- /dev/null +++ b/questions/BounceBall/dataframes/d67bbb7d87d745ad8b5f8871e4ae9a10.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04ef38a80f470c079bf5964464d5a2319e221d0f39b009e1c40fb512ff45f18c +size 52510 diff --git a/questions/BounceBall/dataframes/da62c8d5d796459081533a11ff95d0c9.parquet b/questions/BounceBall/dataframes/da62c8d5d796459081533a11ff95d0c9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bb29796659dbe7cf16a21f406669e35d4615e9ae --- /dev/null +++ b/questions/BounceBall/dataframes/da62c8d5d796459081533a11ff95d0c9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4b962765dc00b7965afa823c97962d52c57ef28e1895d00e94f296766ae97aa +size 52544 diff --git a/questions/BounceBall/dataframes/da754a2fc8be40ee99add59263424786.parquet b/questions/BounceBall/dataframes/da754a2fc8be40ee99add59263424786.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6db739c85192369a06842d8897e06025a90d4262 --- /dev/null +++ b/questions/BounceBall/dataframes/da754a2fc8be40ee99add59263424786.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20435a388826d38ccc1900d791c4deda749c1c845f52f92b05b27ae860dd4fad +size 52510 diff --git a/questions/BounceBall/dataframes/e386c1a131f5492f966c7fd71a723ae9.parquet b/questions/BounceBall/dataframes/e386c1a131f5492f966c7fd71a723ae9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..eb4321a860a9c84bcbf7b673f27f5afd7c973614 --- /dev/null +++ b/questions/BounceBall/dataframes/e386c1a131f5492f966c7fd71a723ae9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44c36ebdfa510e4d634203f48d3c9aeea2446f9e39791ae222a1792cbaa41cf5 +size 52520 diff --git a/questions/BounceBall/dataframes/e4cf1fe8d4165cb1845b624096980de5.parquet b/questions/BounceBall/dataframes/e4cf1fe8d4165cb1845b624096980de5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..30acfd95d0eb0e65525d4ed89f9b3785694e4ec2 --- /dev/null +++ b/questions/BounceBall/dataframes/e4cf1fe8d4165cb1845b624096980de5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee47c7b3b398613c6d5a899156eded97d5fe86d00001f168efe16f08df532bb6 +size 52545 diff --git a/questions/BounceBall/dataframes/e5d0ca629a2e49bcb76945a581cd1a98.parquet b/questions/BounceBall/dataframes/e5d0ca629a2e49bcb76945a581cd1a98.parquet new file mode 100644 index 0000000000000000000000000000000000000000..44c0fe0b98f1b1cf2fba4c83eb71d019b26622f5 --- /dev/null +++ b/questions/BounceBall/dataframes/e5d0ca629a2e49bcb76945a581cd1a98.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f66f1315a5e1a71cc56d9485d284e09d01001720dea0f65abb227080b449f7c2 +size 52544 diff --git a/questions/BounceBall/dataframes/e6614388019a4d22bdd9140403a47547.parquet b/questions/BounceBall/dataframes/e6614388019a4d22bdd9140403a47547.parquet new file mode 100644 index 0000000000000000000000000000000000000000..66e877473c97120d68447816256f11d7099831fa --- /dev/null +++ b/questions/BounceBall/dataframes/e6614388019a4d22bdd9140403a47547.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44f494323bf11c68e0be1789bf1a7ef6174d47478d96ef9a64ec2418a6be6783 +size 52544 diff --git a/questions/BounceBall/dataframes/e6b5a1b16cc15dbe8d1fe3b51eb03100.parquet b/questions/BounceBall/dataframes/e6b5a1b16cc15dbe8d1fe3b51eb03100.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b96cd297d2df0d6f3c18d4a8ebdcae95ccc1a1d8 --- /dev/null +++ b/questions/BounceBall/dataframes/e6b5a1b16cc15dbe8d1fe3b51eb03100.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d19391c0b0894c99d09ed5f79903059a504a4723e5f8661049c6be2a6a1ba474 +size 52670 diff --git a/questions/BounceBall/dataframes/e755a27e73ee5af2bf1337a1a71d0ea3.parquet b/questions/BounceBall/dataframes/e755a27e73ee5af2bf1337a1a71d0ea3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bbeead0dbcb63db900c81df3d85cbaf0638ad8be --- /dev/null +++ b/questions/BounceBall/dataframes/e755a27e73ee5af2bf1337a1a71d0ea3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:510f41ff93a761638fab8f806a0e17161e82e5309801a5e99211662bbc26df85 +size 52520 diff --git a/questions/BounceBall/dataframes/e957c95514144e96beeead74e1ca010a.parquet b/questions/BounceBall/dataframes/e957c95514144e96beeead74e1ca010a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2e47a5c808d9cb1ed173a51420e6ed4bc6e01bc3 --- /dev/null +++ b/questions/BounceBall/dataframes/e957c95514144e96beeead74e1ca010a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d00034baf857018ef6f6ee092b8c5aa7934a4413c2afcb9c5516a0ba0ead6ba0 +size 52520 diff --git a/questions/BounceBall/dataframes/e9cc36d951655bdfbcf90d0bb2322a0e.parquet b/questions/BounceBall/dataframes/e9cc36d951655bdfbcf90d0bb2322a0e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f1822351a8a1fafd157f105fef02a68e076ba69a --- /dev/null +++ b/questions/BounceBall/dataframes/e9cc36d951655bdfbcf90d0bb2322a0e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b6e8aacee77380e20669bc54ea3fafb4019997c32fc79d2384c45e345539ae3 +size 52576 diff --git a/questions/BounceBall/dataframes/ebb3d1ed8560449a8e1113f5f0aee8da.parquet b/questions/BounceBall/dataframes/ebb3d1ed8560449a8e1113f5f0aee8da.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bf420934f1e315d70b7c672d923f28eff5501d23 --- /dev/null +++ b/questions/BounceBall/dataframes/ebb3d1ed8560449a8e1113f5f0aee8da.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7d121f249d1b246467d7bfd4b4d67fbda373c66786ca34103bc9d46cc1e6eae +size 52520 diff --git a/questions/BounceBall/dataframes/ed3abb63ca7e49dca5dbc6f41691f421.parquet b/questions/BounceBall/dataframes/ed3abb63ca7e49dca5dbc6f41691f421.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0476d27c8dad84f47c1629e79978ed6955fb78cb --- /dev/null +++ b/questions/BounceBall/dataframes/ed3abb63ca7e49dca5dbc6f41691f421.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:708ba99c7645a91fd8abea944e88d6a4854929056ce51343f41f862979a291fd +size 52534 diff --git a/questions/BounceBall/dataframes/f3359983559e491da7139fb1d378f667.parquet b/questions/BounceBall/dataframes/f3359983559e491da7139fb1d378f667.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9ef5a6ca1bd86193d3922fb26c52010626c775ca --- /dev/null +++ b/questions/BounceBall/dataframes/f3359983559e491da7139fb1d378f667.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b436086178cbae720523f30925e2189afc12ab095e58b0cf2919f1ece03e456 +size 52520 diff --git a/questions/BounceBall/dataframes/f9bf7b07efb646ecbc9d989085c45da8.parquet b/questions/BounceBall/dataframes/f9bf7b07efb646ecbc9d989085c45da8.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ec0d81ac419fd976a9f05dc767e07b8203097091 --- /dev/null +++ b/questions/BounceBall/dataframes/f9bf7b07efb646ecbc9d989085c45da8.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94f7ecd271ee439510fe7f5f821e6fa77d29faa16af645106a90d3f101c2e984 +size 52510 diff --git a/questions/BounceBall/dataframes/fab4f5f11b874ef58e909995690d57ec.parquet b/questions/BounceBall/dataframes/fab4f5f11b874ef58e909995690d57ec.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2cecc170b7ffd05adc64cf01a269906320533201 --- /dev/null +++ b/questions/BounceBall/dataframes/fab4f5f11b874ef58e909995690d57ec.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c0ec9604cbee2f99476f3695b15b0a6fe1d11d82d1854c437a8471daad3992c +size 51377 diff --git a/questions/BounceBall/dataframes/fc653cb5b3c05a259ddecd3547667a15.parquet b/questions/BounceBall/dataframes/fc653cb5b3c05a259ddecd3547667a15.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b208cdb0311799237918c3d61bd52716d7a73489 --- /dev/null +++ b/questions/BounceBall/dataframes/fc653cb5b3c05a259ddecd3547667a15.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba73ea8a1a1aaf68b477599230bbfd16a010412502de906037502800d316569d +size 52520 diff --git a/questions/BounceBall/model_record.json b/questions/BounceBall/model_record.json new file mode 100644 index 0000000000000000000000000000000000000000..139b12fb9bdfbdabbf28102745b37d478a0471d4 --- /dev/null +++ b/questions/BounceBall/model_record.json @@ -0,0 +1,7499 @@ +{ + "35ce9600f2b649f39794bad12c8451f5": { + "parameters_hash": "b054165364e7ab8827bbcaf351c56534", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T00:04:17.361771", + "end_time_simulation": 59.97999954223633 + }, + "e4b5230c7f30481aaf9c63f92b15fd6c": { + "parameters_hash": "f4595e0b4640262b39fb2592e46032e3", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T00:04:23.403205", + "end_time_simulation": 59.97999954223633 + }, + "ecba22c2157541408c5a4f9cb7ba2841": { + "parameters_hash": "d7e36c0cf3a9ff1ea0a5d141a1d96c4e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:04:26.894790", + "end_time_simulation": 59.97999954223633 + }, + "5bd91aea189940ae8b8e749e0bff99f0": { + "parameters_hash": "c72e917404eaf0341864b875286b1ccb", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T00:04:30.219164", + "end_time_simulation": 59.97999954223633 + }, + "9754fa1fec5b5149856c4639b3a143ba": { + "parameters_hash": "408e97d95e04dc4e7bc3c73e20f3a6d5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:04:33.412300", + "end_time_simulation": 59.97999954223633 + }, + "dd9f4ab010164d7e8a8f0f92877761bc": { + "parameters_hash": "68aa6347e58dc1f5b4e75ccbf803a078", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-02T00:04:36.349579", + "end_time_simulation": 59.97999954223633 + }, + "16f4a46042574897b22a6cfc4b38a707": { + "parameters_hash": "50e452ca5226bf6b902e95d9d0b2dae1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:04:39.235793", + "end_time_simulation": 59.97999954223633 + }, + "3731111840514a3a84fa9d182c7fb512": { + "parameters_hash": "d70d86fcdb1ed345b6cf75172957e4f7", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-02T00:04:42.414813", + "end_time_simulation": 59.97999954223633 + }, + "fe354cf7c45652f49b1c6fa03d959400": { + "parameters_hash": "8e8ba14a592222fa4868c0a51a238bd4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:04:45.372078", + "end_time_simulation": 59.97999954223633 + }, + "ddd6fbf52768417e9291d4bbc26742b9": { + "parameters_hash": "26dec4dbbfa0a04855c74475d3ff1223", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-02T00:04:48.608431", + "end_time_simulation": 59.97999954223633 + }, + "72dcd97718cb50178861570db36814f5": { + "parameters_hash": "4d4687f6a1c1338a1209381b436d9959", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:04:51.577687", + "end_time_simulation": 59.97999954223633 + }, + "d217e2d448755e4bb51130635edf22b9": { + "parameters_hash": "a711d5ad4b1191cbb8bf4e986e311740", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-02T00:04:54.872367", + "end_time_simulation": 59.97999954223633 + }, + "1070b1855d2d5ab681b8d3acda8383ba": { + "parameters_hash": "f0f1a170739e3effe95371a6f4642b52", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:04:57.800618", + "end_time_simulation": 59.97999954223633 + }, + "7b377fd33ec55476b4dd35ce0f6adf5b": { + "parameters_hash": "f9455494e70645b8c2ea7aba858e9449", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-02T00:05:01.353460", + "end_time_simulation": 59.97999954223633 + }, + "9ef5dcb5a3d25f0791c64c8e9fe64fd1": { + "parameters_hash": "43ad52789d6a06019f41f8b5a7846213", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:05:04.305733", + "end_time_simulation": 59.97999954223633 + }, + "55cd6c88ee704897a0318d3680135fb7": { + "parameters_hash": "8c8bedd0a9bb68365bf480743ddb2721", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:50:49.520151", + "end_time_simulation": 59.97999954223633 + }, + "5f2cb0f3b6105e8cb9ea7dc0aad12eab": { + "parameters_hash": "cd4c831c753752a5e6cd2bfaa8509cfc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:50:52.693060", + "end_time_simulation": 59.97999954223633 + }, + "b535149375454af8a7c85d6e3906aa55": { + "parameters_hash": "68e9195f1c826774552a920a2870dcd1", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:50:56.107803", + "end_time_simulation": 59.97999954223633 + }, + "9f718d42abe659e9ab19483717316ef2": { + "parameters_hash": "4fe62060b9eb75030225229852ae50c0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:50:58.488677", + "end_time_simulation": 59.97999954223633 + }, + "f783b15cd8904a02838281c2d01da80d": { + "parameters_hash": "65a8feecdc3b876963194eb526c33c45", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T23:56:51.207836", + "end_time_simulation": 59.97999954223633 + }, + "06b63b21dde051af8ab9e70b3f28559b": { + "parameters_hash": "313d83865deecfcf5d315c45b652236a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:56:53.946205", + "end_time_simulation": 59.97999954223633 + }, + "81064a33c383447293b0d82c15f891bb": { + "parameters_hash": "a25869b63aeea4d24d4b36d2ffb62d10", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T23:56:56.924743", + "end_time_simulation": 59.97999954223633 + }, + "8e7cff6dc5055dd88e41d0f827887322": { + "parameters_hash": "895346f1887866231655f2a268d22d64", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:56:59.715881", + "end_time_simulation": 59.97999954223633 + }, + "e386c1a131f5492f966c7fd71a723ae9": { + "parameters_hash": "92850292fd8bd6a522c476d1ce6b14d7", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T13:44:47.226427", + "end_time_simulation": 59.97999954223633 + }, + "1aac2ed4a25544c894c596067aeb18ae": { + "parameters_hash": "841e717955365a497a1f405c126bd395", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T13:44:52.786379", + "end_time_simulation": 59.97999954223633 + }, + "a4f1843b905c43f69951347df4f7f483": { + "parameters_hash": "bb485c251f2046c527dbc89b7411300c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:44:57.006240", + "end_time_simulation": 59.97999954223633 + }, + "f5cd4a6d504c45609acb1b357f728bd2": { + "parameters_hash": "dc1c6bc4baca84e6c486d5ec9a65987f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T13:45:00.807682", + "end_time_simulation": 59.97999954223633 + }, + "1d8e080f64ae5bdeb8485e4a48ecbb58": { + "parameters_hash": "0a7ad4224fcb3d29f4ef40726fceb8af", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:45:03.806365", + "end_time_simulation": 59.97999954223633 + }, + "f6d835a5b86e4ec6920796e0f615111d": { + "parameters_hash": "67be57d62066176b7aaec3a1d1ea9727", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T13:45:07.388646", + "end_time_simulation": 59.97999954223633 + }, + "fa843c0f168e4593ac2f76121188e681": { + "parameters_hash": "2e2903cc16d815254436fed191d8b72d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:45:10.790399", + "end_time_simulation": 59.97999954223633 + }, + "5f6f4011eba24b46a09cac472f54fff4": { + "parameters_hash": "378ac60f3f79eac1e573015cfe1d5d38", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T13:45:14.351901", + "end_time_simulation": 59.97999954223633 + }, + "d27ecf7dd7ed492c97a16edbc646abb6": { + "parameters_hash": "7e0aac4bf4e82d7a360c419b405397d1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:45:17.251050", + "end_time_simulation": 59.97999954223633 + }, + "1a6d28e85a7a41a292e5c95ea91ee2cb": { + "parameters_hash": "365ec0182691f380b069383e4df9861e", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:48:00.430310", + "end_time_simulation": 59.97999954223633 + }, + "8c9de2218ac34cfc8539019e89e2fd3f": { + "parameters_hash": "bd983ad70af70333439952032972d861", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:03.856366", + "end_time_simulation": 59.97999954223633 + }, + "8f1a15c7ef054d7f88d29ec91f1620f9": { + "parameters_hash": "d5c6f0c8530f26d2ea2036eb9c200e9a", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-04T13:45:26.912751", + "end_time_simulation": 59.97999954223633 + }, + "9a53680e4f0a40718d33337c63544fef": { + "parameters_hash": "344e3f71f7e3f027fad167dd70c7c183", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:45:29.847432", + "end_time_simulation": 59.97999954223633 + }, + "c4bb9e9601ed5118837aa51f134d44b1": { + "parameters_hash": "8ccf06ebc0774359e5d42365191cb8a5", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T13:45:32.808628", + "end_time_simulation": 59.97999954223633 + }, + "0fe4728047355e1d81700973a492f4e5": { + "parameters_hash": "a52d96a799936fa0c4b317221164146a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:45:36.129506", + "end_time_simulation": 59.97999954223633 + }, + "bc1476a263f35ad0859eef0c3a514efb": { + "parameters_hash": "7c164d7eaecd29f7d917b82c93f36ee9", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T13:45:39.484430", + "end_time_simulation": 59.97999954223633 + }, + "cd7d4173b85a5adb9deb28c6d5418590": { + "parameters_hash": "44679b1942921cda7ea4e4c6d249d82e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:45:42.869435", + "end_time_simulation": 59.97999954223633 + }, + "e957c95514144e96beeead74e1ca010a": { + "parameters_hash": "34c00222d6beb05c3f71a4a79887107e", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:01.341737", + "end_time_simulation": 59.97999954223633 + }, + "bf8f2a6e51f955a79d815f4ce9e97d1a": { + "parameters_hash": "8641c509b9480ddbbcb7c175e1c46741", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:03.842840", + "end_time_simulation": 59.97999954223633 + }, + "7bb11a8651b44883857ac1b7ed642e6a": { + "parameters_hash": "dfea1896ef1cfd4acc942e4b0d2b77ed", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:06.657964", + "end_time_simulation": 59.97999954223633 + }, + "55867e0cb0345aec95040b137cccf16e": { + "parameters_hash": "a515f6da72fff44cf95d2e506fdfd3b3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:08.844568", + "end_time_simulation": 59.97999954223633 + }, + "be12a2ede7744ecdae83f9e0d1f2a47f": { + "parameters_hash": "a682076279644b3c16f136b1e48eef3a", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T23:57:14.437275", + "end_time_simulation": 59.97999954223633 + }, + "5e627fe0660b5bd5b201e7b9df793a77": { + "parameters_hash": "f2f4313cd444e38949ea70f52f33056e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:57:17.410829", + "end_time_simulation": 59.97999954223633 + }, + "664fa7515eea439692b7f460afbf1631": { + "parameters_hash": "26ae001a83106a1ff76cc174e0797c70", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T23:57:20.359100", + "end_time_simulation": 59.97999954223633 + }, + "20075e2a0c725b60b09466e556d456aa": { + "parameters_hash": "c1f79cf7c186144669c152f9ac370872", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:57:23.635592", + "end_time_simulation": 59.97999954223633 + }, + "bbe78dcf65cf4990b574beb1df877cfc": { + "parameters_hash": "09897ffa321559f9ad73838b55cbba8d", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T19:41:40.571482", + "end_time_simulation": 59.97999954223633 + }, + "98ce3759ae534c9f8ee938e151b01032": { + "parameters_hash": "e9eca74cd5d96b1881f09c2085afb794", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:41:46.086171", + "end_time_simulation": 59.97999954223633 + }, + "1415b36e72464064bd01d4a65b8e4894": { + "parameters_hash": "b805647631db5d6bfbdec5eb1902158f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:41:49.486807", + "end_time_simulation": 59.97999954223633 + }, + "491d9b6c80eb410d9716acd5fe7837a1": { + "parameters_hash": "93aee41269bf95a90cdc9518a7258a17", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:41:53.173926", + "end_time_simulation": 59.97999954223633 + }, + "ae1437a5a4e55893882ca57a61d7dbdf": { + "parameters_hash": "1c11e013d26d3dbcadd34908cc4d5fd6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:41:56.640553", + "end_time_simulation": 59.97999954223633 + }, + "fda05117e7884597920df1a772600087": { + "parameters_hash": "4db9373d033d2cb3b51c47b87395b401", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T19:42:00.202156", + "end_time_simulation": 59.97999954223633 + }, + "4f361121267b40d49afd857c55898026": { + "parameters_hash": "d486686affcea5d9d94cdb930de5aa91", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:03.221824", + "end_time_simulation": 59.97999954223633 + }, + "5e2085aaa72d4790a04a0c06ea42b3eb": { + "parameters_hash": "21e28aaf0b4044c9f75a15febb1b163a", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T19:42:06.266546", + "end_time_simulation": 59.97999954223633 + }, + "cbcc3939903b43e6a4da52e00ef9e5ee": { + "parameters_hash": "f58fdd3333f49800def99aa8eac6b798", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:09.068234", + "end_time_simulation": 59.97999954223633 + }, + "12dd1c92dc4649d4808421b7c2dbb124": { + "parameters_hash": "671dc61016016609cc90760b4d2b9013", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T19:42:12.202117", + "end_time_simulation": 59.97999954223633 + }, + "44112a12e9a349bea746cdab059a8ddf": { + "parameters_hash": "384e8d0264f5e6057fcdb98961a580e2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:14.885201", + "end_time_simulation": 59.97999954223633 + }, + "255c6fa1abd143f79b0fb957898bc1e8": { + "parameters_hash": "294974a78b6ca5ac57f0301be790c46f", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:42:18.280534", + "end_time_simulation": 59.97999954223633 + }, + "3a976810870b460aa067f749bac9c034": { + "parameters_hash": "6d44ba4032290d7fd3a39f6bf02d1102", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:21.302272", + "end_time_simulation": 59.97999954223633 + }, + "ab3018e866055b7d86378f776ed18b5a": { + "parameters_hash": "ed30362f9bdca29993688420bc2d5234", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:42:25.067402", + "end_time_simulation": 59.97999954223633 + }, + "336d7a7c5ae051cd8157146213bc9be5": { + "parameters_hash": "ba587c81cb012cc1d5845c64a2ba5ce8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:27.377330", + "end_time_simulation": 59.97999954223633 + }, + "e2130b83970f5def9a9f81547bf78d8e": { + "parameters_hash": "c487960d48ef84532d3dafcb1ba67099", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:42:29.753552", + "end_time_simulation": 59.97999954223633 + }, + "d797922581b550239d6d70e84140b6bb": { + "parameters_hash": "bd7438c15f7590b188df8a56bac501a7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:32.976580", + "end_time_simulation": 59.97999954223633 + }, + "f3359983559e491da7139fb1d378f667": { + "parameters_hash": "b7877b6e765cb0b52a88931e79f15116", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:11.615411", + "end_time_simulation": 59.97999954223633 + }, + "879ffcd40ec95b0281c3f80a55fa8eb1": { + "parameters_hash": "a675cb9b5f233bfe7ea7e9f7d6f91125", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:13.941982", + "end_time_simulation": 59.97999954223633 + }, + "2afe39e081a14923a1d46c052479dc3e": { + "parameters_hash": "484e10d4b2e22c945701c5abb5e96c53", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:16.656361", + "end_time_simulation": 59.97999954223633 + }, + "5514be07a2c65f1dbea6798d3c7e063b": { + "parameters_hash": "02a7ad6e1b6c7d61ea49fba49e354dda", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:19.531756", + "end_time_simulation": 59.97999954223633 + }, + "ebb3d1ed8560449a8e1113f5f0aee8da": { + "parameters_hash": "76f91b400bf19af2e7734c0a010fe808", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T23:57:31.947122", + "end_time_simulation": 59.97999954223633 + }, + "94787f272af0526a8d401dc7ace111dd": { + "parameters_hash": "aff542e912c3b6e5f7982ccf58c4b02c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:57:34.621663", + "end_time_simulation": 59.97999954223633 + }, + "4e53c84aa5024bf4ae26bb23d7a4d07e": { + "parameters_hash": "035d38b8be32e0e01aab8449ac8c23d3", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T23:57:37.091804", + "end_time_simulation": 59.97999954223633 + }, + "f70b83b6567f5f4baa47db2c7aa33649": { + "parameters_hash": "00e0a8f0f7808cb175432c8d415d5ee1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:57:39.474553", + "end_time_simulation": 59.97999954223633 + }, + "22bd886147e847e9a0420134bbfcd2c8": { + "parameters_hash": "1eb602b03135fca1ba4be3c40f5ed5bf", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T23:57:41.817883", + "end_time_simulation": 59.97999954223633 + }, + "029634bc34895867af8d85b4133bfdd4": { + "parameters_hash": "c4dec0d6b44a9a7c91f74dd83f7810ba", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:57:44.595420", + "end_time_simulation": 59.97999954223633 + }, + "a17a6c0b50f444b380cd3867d55fee17": { + "parameters_hash": "edb882af8931a41ce5a394599798434f", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T14:22:23.264789", + "end_time_simulation": 59.97999954223633 + }, + "5990d64f12064184ac3587e5004bac94": { + "parameters_hash": "9979a4c1ee404b9a914673b8429008de", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T14:22:28.631182", + "end_time_simulation": 59.97999954223633 + }, + "2b6a3b2757344b48a7d4848de5678ca4": { + "parameters_hash": "5c3b04a586a515af6fa5be4ac58447a9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T14:22:32.726284", + "end_time_simulation": 59.97999954223633 + }, + "621434034718493d90ba0fab2e110fd6": { + "parameters_hash": "4d3768cefc07b4496e6430a643e6b50c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T14:22:36.876668", + "end_time_simulation": 59.97999954223633 + }, + "69154da064015e1b9b93b021e9a4e71c": { + "parameters_hash": "daaa9a9abf245c8c5b7508b70f9d09f2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T14:22:40.211855", + "end_time_simulation": 59.97999954223633 + }, + "13862d4768c54163b576c9c0495a186a": { + "parameters_hash": "2123eea18e335e624cced7d29f989b77", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T14:22:43.623358", + "end_time_simulation": 59.97999954223633 + }, + "bead021b7e9e4d81b32c8fc94764ba78": { + "parameters_hash": "1ec70b9eda0bc35808bc3ba452f8e583", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T14:22:47.030815", + "end_time_simulation": 59.97999954223633 + }, + "7dcddce7cc2a472a91c22cfe7b61f906": { + "parameters_hash": "93ef51a99f4d68a0ec1363409f392e92", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T14:22:50.735338", + "end_time_simulation": 59.97999954223633 + }, + "fd844614763542a19ea03d870729e741": { + "parameters_hash": "4e1445ad736192d7c8ccc615386a94b4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T14:22:53.966361", + "end_time_simulation": 59.97999954223633 + }, + "afafe73e408e404a96c1311dcc2562a8": { + "parameters_hash": "d90c0dd50d351f541877816de1871071", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:48:07.644153", + "end_time_simulation": 59.97999954223633 + }, + "89f51b0a47eb482badc28415e27573a6": { + "parameters_hash": "634d5b9f26e298808d49eb78346989a6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:10.604642", + "end_time_simulation": 59.97999954223633 + }, + "90674e94122149fb8dc50ad9819d0559": { + "parameters_hash": "eb76b7dec909ccfa2aae65d9a0abc7ac", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T22:18:27.554752", + "end_time_simulation": 59.97999954223633 + }, + "e97751b9dbb35d4a8e3e22f6c85a0509": { + "parameters_hash": "b0a1f0fd8be843fe4e5ec66249b8ea62", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T22:18:30.020883", + "end_time_simulation": 59.97999954223633 + }, + "d43b4117fdc8568e89abe3cbb270350f": { + "parameters_hash": "1c2448c65b78a7752f8302d26115ed30", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T14:23:03.644073", + "end_time_simulation": 59.97999954223633 + }, + "f901364ac01752ecb91dfa37464dc2da": { + "parameters_hash": "8cebb9bab73fca45837338e7512bd02b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T14:23:06.932881", + "end_time_simulation": 59.97999954223633 + }, + "3fd9c7f94e355f65bada44cc585647d6": { + "parameters_hash": "ee0fc1ebb0b9837152c273192bf66319", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T14:23:10.279557", + "end_time_simulation": 59.97999954223633 + }, + "bfd66cd6dc8e500ea612cdf00e4b3816": { + "parameters_hash": "7948d225a519af53b0110772f9a01d5e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T14:23:13.261764", + "end_time_simulation": 59.97999954223633 + }, + "f9bf7b07efb646ecbc9d989085c45da8": { + "parameters_hash": "df68d53d89636da356cfa7e64eb41e90", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:22.467205", + "end_time_simulation": 59.97999954223633 + }, + "3307c9b6154452b395b7eae7e3d793eb": { + "parameters_hash": "037ad6c60497f6e2fa1f8f63810627ab", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:25.042726", + "end_time_simulation": 59.97999954223633 + }, + "3ca82ae3c7a844c08d9b777385bcf923": { + "parameters_hash": "01b8f83085d3b85b6beda89beb9541c6", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:27.671651", + "end_time_simulation": 59.97999954223633 + }, + "e99421557f4b5d49b68cf75b38842d40": { + "parameters_hash": "84deb0d5b75881374be93ae3113a326c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:29.994006", + "end_time_simulation": 59.97999954223633 + }, + "4483b5b67b3f4d9abae899f3bea83fa0": { + "parameters_hash": "5a4e3bfbfe29a777450556dee062d92f", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T23:58:07.137610", + "end_time_simulation": 59.97999954223633 + }, + "42f249109059504d9357381fdbfc35ea": { + "parameters_hash": "5e7ba494f0d67103cb428bd6ddad5db5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:58:10.365219", + "end_time_simulation": 59.97999954223633 + }, + "6783ef519fdc4d7bb71f66db00c264e7": { + "parameters_hash": "6313b22ddcd216e342cf188494d40bbc", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T14:02:58.124094", + "end_time_simulation": 59.97999954223633 + }, + "ae74a21b292349da931ebbc527ad8729": { + "parameters_hash": "b0213d150101563b8e56ea042aa76f18", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:15:24.347221", + "end_time_simulation": 59.97999954223633 + }, + "1d110ce95ad64424b2f478fc324adc63": { + "parameters_hash": "5f3146c9c28d7f68f023858049902773", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:15:27.453394", + "end_time_simulation": 59.97999954223633 + }, + "e000e5879c124ed0bcb231945053ec6f": { + "parameters_hash": "554e417801c1ef72cd55b088eaddf533", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:15:30.720450", + "end_time_simulation": 59.97999954223633 + }, + "3ca1ea499b2456b8bf314ebaf952957b": { + "parameters_hash": "493ce250665909b1cb8c867fb418fac7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:15:33.159266", + "end_time_simulation": 59.97999954223633 + }, + "6a1ab4ced271460588b9fd1071498d96": { + "parameters_hash": "aeaf4f5c24cace1bf2fd33a8358a5c14", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:15:35.892273", + "end_time_simulation": 59.97999954223633 + }, + "6838d6a48c264a6aa9dc9c3cf3b4f48f": { + "parameters_hash": "5b6aeae16c2edcb71529ae84f3b35b41", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:15:38.309655", + "end_time_simulation": 59.97999954223633 + }, + "6c4ff7108da34c2d94ba2bb300dd1ee7": { + "parameters_hash": "6a0550716d5eefc02e5a3d0ba2a08589", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:15:41.026316", + "end_time_simulation": 59.97999954223633 + }, + "402d1cee8e3c43af88d0a13c0b18ffd6": { + "parameters_hash": "d74836804b48b60ed523b4085c5d80ec", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:15:43.224417", + "end_time_simulation": 59.97999954223633 + }, + "245d11d70caf412eb50b80db74cfb5a0": { + "parameters_hash": "f68a5f99d2430789417bfec590c86048", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:15:46.020936", + "end_time_simulation": 59.97999954223633 + }, + "a4517575501d459aaef4a493caad4aa1": { + "parameters_hash": "171b4e21a3646d8b276c9839947f62e7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:15:48.396103", + "end_time_simulation": 59.97999954223633 + }, + "4f26efbd7d274f7fa043588a6d4e7204": { + "parameters_hash": "64f3ddce27f31b9286f3e236b7aec0fe", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:15:50.918920", + "end_time_simulation": 59.97999954223633 + }, + "dc78a5c8606349e2bd7999f088e22796": { + "parameters_hash": "127afc42ca02742bfb64059861163979", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:15:53.304683", + "end_time_simulation": 59.97999954223633 + }, + "504d1a083a345050bd1434ebcf06dc11": { + "parameters_hash": "ec0b5c3bea935b677701fb85a6c35f80", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:15:56.027448", + "end_time_simulation": 59.97999954223633 + }, + "6652076bfb3e5b739998b09ebf05a8e6": { + "parameters_hash": "86dee074f2650ca3c3df32bec89132ed", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:15:58.505956", + "end_time_simulation": 59.97999954223633 + }, + "e6b5a1b16cc15dbe8d1fe3b51eb03100": { + "parameters_hash": "6028a70fcdc33ddaf14ddd1e660f34d3", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:16:01.203472", + "end_time_simulation": 59.97999954223633 + }, + "49c04867c6ed5e7e91ecd8e5d6be4d52": { + "parameters_hash": "ae3a05c74521fc65337a4d4e7f591afa", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:03.982853", + "end_time_simulation": 59.97999954223633 + }, + "5f71b0ed2746429783aad22c75fac520": { + "parameters_hash": "ba6015f7b64de3fb657d01c872205c96", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:16:06.790670", + "end_time_simulation": 59.97999954223633 + }, + "04596884bda052829ca3b5a2896421a2": { + "parameters_hash": "613b25796a9e59d6ca9153ab07b7aaf1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:09.258730", + "end_time_simulation": 59.97999954223633 + }, + "fb3fa55a92ac477f82701ae9f31eb889": { + "parameters_hash": "3e1232e55e09ea9c9705c32bc6cd4d16", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:16:12.372808", + "end_time_simulation": 59.97999954223633 + }, + "f0030209b3cf5ee9b202b2ebba6585fa": { + "parameters_hash": "ab10bdc33fae875127e188383a0aff7b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:14.657047", + "end_time_simulation": 59.97999954223633 + }, + "c76b02db082640a5b38a94d575d78138": { + "parameters_hash": "15d90497dc23bc8c817fee522cfa16bb", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:16:17.505325", + "end_time_simulation": 59.97999954223633 + }, + "5774180fe0535fd88512cf176bcb27ff": { + "parameters_hash": "a3af9b99a72be2bbfb9c8815094e6df8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:19.869828", + "end_time_simulation": 59.97999954223633 + }, + "c1de5671632d441fbd2db5677b9cb020": { + "parameters_hash": "3e3d922484db0f6ac4d72469b0bf925b", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:16:22.574209", + "end_time_simulation": 59.97999954223633 + }, + "aec20ae0964b52e3ad40583b898ffc5c": { + "parameters_hash": "9b3677286f7a957d3293119ce4d7606b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:25.120524", + "end_time_simulation": 59.97999954223633 + }, + "84bbd6c06ddb4b54a3f12c882c2eccc4": { + "parameters_hash": "6adebb0e7c5af169480e03a11d7cab26", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-06T00:16:27.578395", + "end_time_simulation": 59.97999954223633 + }, + "7dacaf46a9ec45cf9c7e1aabad31e83f": { + "parameters_hash": "9eff31f3977f47d94aee7b8d1c72570c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:16:30.110333", + "end_time_simulation": 59.97999954223633 + }, + "867d2bc547324ea8ac75ed6e2562e467": { + "parameters_hash": "48cd9dbdbf49bf9bac0c3f8d000eea01", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:32.560550", + "end_time_simulation": 59.97999954223633 + }, + "0489ea21a5664c8f99c10fb365fac893": { + "parameters_hash": "66f1109a9d9f58799719440154ca75dc", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:16:35.241151", + "end_time_simulation": 59.97999954223633 + }, + "757b8e3ac07357c2a6d310ca00073d45": { + "parameters_hash": "ed1f259a3e6d86dcf124a175e6285a7d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:37.803066", + "end_time_simulation": 59.97999954223633 + }, + "7db96ce9ad0240a28b2a65383312bbb0": { + "parameters_hash": "62c6cf91a04f8eacb16e765375ec8547", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:16:40.710872", + "end_time_simulation": 59.97999954223633 + }, + "f204790b84cb4927805d665a559ddd58": { + "parameters_hash": "55899428bd04289dd4e164c61a97658f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:43.190510", + "end_time_simulation": 59.97999954223633 + }, + "1b364086c3c740789d832cfebdd6cb1e": { + "parameters_hash": "3fca2198a402cddcc08c6f711ddd1ba5", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:16:45.668574", + "end_time_simulation": 59.97999954223633 + }, + "cd666a926a6e4994abc4627695f1fcee": { + "parameters_hash": "191f8e756d9ad772f5e597fc622f587b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:48.103329", + "end_time_simulation": 59.97999954223633 + }, + "35428ae77661442ab6f33894965097ef": { + "parameters_hash": "dc6ab5bb524c1b40c954247a4f33e6c5", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:16:50.795362", + "end_time_simulation": 59.97999954223633 + }, + "e05257c017934f898486f0d1f5a22605": { + "parameters_hash": "b76f292d9d768110e749ed8cce8cab74", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:53.102056", + "end_time_simulation": 59.97999954223633 + }, + "4d8fb279215248f8acbfad4f5e395b2e": { + "parameters_hash": "11c88f49a40db5a915e37952eb0d7053", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:16:55.869057", + "end_time_simulation": 59.97999954223633 + }, + "9a9992b789e951fd9d29ae035d5da63d": { + "parameters_hash": "d87325216453b4a2f5aefd58b422f3b4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:16:58.369690", + "end_time_simulation": 59.97999954223633 + }, + "73e57408da7a53bba190eae95592a58b": { + "parameters_hash": "9dad8771ba3f0ce58f7a85d79f7af2d6", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:17:01.223196", + "end_time_simulation": 59.97999954223633 + }, + "ace41999ea44558882cc9797cf8d66a5": { + "parameters_hash": "2cfabdec37a2dff917adbfcaad03c997", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:17:03.791171", + "end_time_simulation": 59.97999954223633 + }, + "b834f44627dc54bf844fbe7d411d8468": { + "parameters_hash": "adbafecdf5bb19ae328913e52cd979ed", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:17:06.638554", + "end_time_simulation": 59.97999954223633 + }, + "9fe863b05a585e1cbde5c70ee27a7d1c": { + "parameters_hash": "023da36ce7a1ffc1c5a08200340d3fe9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:17:08.908377", + "end_time_simulation": 59.97999954223633 + }, + "c7647d63ffa142a181f3bc6d78795879": { + "parameters_hash": "2d1bf83f69cc4a383d75275447f73aa8", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:45:08.736959", + "end_time_simulation": 59.97999954223633 + }, + "f6433388ddbe4dd4ac25a28ffdd659b7": { + "parameters_hash": "c2d4754400e153479188a33b166eb335", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:45:10.913136", + "end_time_simulation": 59.97999954223633 + }, + "cd6bf34b91855921b0b7fbe5d90f35cf": { + "parameters_hash": "08aaf1341aaf1070e068defbc2abeec5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:45:13.046743", + "end_time_simulation": 59.97999954223633 + }, + "da754a2fc8be40ee99add59263424786": { + "parameters_hash": "1781e6f13a65e0fd53f58dcf11718fa9", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:45:15.264413", + "end_time_simulation": 59.97999954223633 + }, + "0818ca7616f54c9a83a8624b17471ec0": { + "parameters_hash": "6c15143db539b060d736a2d179d3ac73", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:45:17.271776", + "end_time_simulation": 59.97999954223633 + }, + "51142e4537fb404aa1f1a4382c92cc17": { + "parameters_hash": "6f7c26919ccd760cd431ded8df85b0ec", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:45:19.603494", + "end_time_simulation": 59.97999954223633 + }, + "7a60e3430e4841feb35ee76c982fd257": { + "parameters_hash": "2f8abc3be4ddc28822832a21978cf297", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:45:21.780114", + "end_time_simulation": 59.97999954223633 + }, + "29a0305a7d8947b0983ab26ff8f5b47a": { + "parameters_hash": "fd7917ec7445186cf3174db36868cb62", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:45:24.000316", + "end_time_simulation": 59.97999954223633 + }, + "fd01749ef17d4eefb7cb66267be43874": { + "parameters_hash": "f7598f779f76b71f6f250fe67e2345cc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:45:25.966539", + "end_time_simulation": 59.97999954223633 + }, + "2888456b9ece4be3bd11a8cc7dbf349b": { + "parameters_hash": "2b7e346a96d4715c8f02067c623fdc62", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:45:28.091740", + "end_time_simulation": 59.97999954223633 + }, + "20018a56d77b473db840ce895671ff29": { + "parameters_hash": "114344064e1897542a290902d867749d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:45:30.293176", + "end_time_simulation": 59.97999954223633 + }, + "367f02720fd2588fb67bd00cbd8ee26e": { + "parameters_hash": "92d9a88d3377377ab79fc9e1a34bcbcb", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:45:32.470335", + "end_time_simulation": 59.97999954223633 + }, + "ab5fa4ad4def5e899e583feeddd58e01": { + "parameters_hash": "101629baf49ab539a0261c3b82644148", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:45:34.498726", + "end_time_simulation": 59.97999954223633 + }, + "78553d66439e5f678fe0b5574efa7384": { + "parameters_hash": "a0c2a34012ee1c36abfc33b74e72aea0", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:45:36.672073", + "end_time_simulation": 59.97999954223633 + }, + "ab05eb6c7d6a5207a77aa8210d5a3237": { + "parameters_hash": "3702fd321ae46156277e4d731e870855", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:45:38.798603", + "end_time_simulation": 59.97999954223633 + }, + "c4c8f3b1864a415d854a5b0f03f8ea71": { + "parameters_hash": "8bc7050c20cb9376498d03961d6d1488", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:42.864675", + "end_time_simulation": 59.97999954223633 + }, + "9fa2d605ae73569f9a91b83cd6a3e02e": { + "parameters_hash": "c9471466f83d9185996b6e875c1ed069", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:45.302648", + "end_time_simulation": 59.97999954223633 + }, + "65f2d269a44d4fffb40ab2833c272d98": { + "parameters_hash": "c159d467f8974f44a125f3b42cd8cd5c", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:51:47.876815", + "end_time_simulation": 59.97999954223633 + }, + "9f42103e7a5d5f5a970d3079777d5125": { + "parameters_hash": "3c590d0a2e89905cb18e10f958283334", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:51:50.531144", + "end_time_simulation": 59.97999954223633 + }, + "4bdd20ee31ce447eb8b038067bdd66d7": { + "parameters_hash": "c61b35e73ff1027e36949073efc0f36f", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T23:58:42.744007", + "end_time_simulation": 59.97999954223633 + }, + "de5683d72a1251f5aa8068f0db994429": { + "parameters_hash": "9d2c187ace0e0bc334c6e8253b83b88f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:58:45.699864", + "end_time_simulation": 59.97999954223633 + }, + "d67bbb7d87d745ad8b5f8871e4ae9a10": { + "parameters_hash": "75de90efb4b12116d19217bae9278b00", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T23:58:48.698702", + "end_time_simulation": 59.97999954223633 + }, + "6b5f8d38a7b0504db11c2fa8088753c6": { + "parameters_hash": "02c23a1801e52896c4822f915d32b150", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:58:51.540348", + "end_time_simulation": 59.97999954223633 + }, + "51a0bffa4af14684ae0f3a57cb35c836": { + "parameters_hash": "3a24db9222a6a9f85e0d089a2f5caaa9", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T23:58:54.200348", + "end_time_simulation": 59.97999954223633 + }, + "eda8470f4c0158dd9936d64f060c1503": { + "parameters_hash": "853082860c643ae85d5772ff1efca887", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:58:56.679968", + "end_time_simulation": 59.97999954223633 + }, + "11e33a2095244285ad1ae8d000034bf2": { + "parameters_hash": "9199d41302f5567aed152b1f3201db49", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-06T00:26:35.176639", + "end_time_simulation": 59.97999954223633 + }, + "07f55a16d69449fcb096b003f6593691": { + "parameters_hash": "82d6de46d567d7269e33254ae97e5438", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:26:39.998716", + "end_time_simulation": 59.97999954223633 + }, + "fac4cfbf90ae4c8eb9d9fcd02861c22c": { + "parameters_hash": "8692134ca51d7ca2d698dc0469938558", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:26:43.134844", + "end_time_simulation": 59.97999954223633 + }, + "86eec73fda0e47d5b2d859782bee9934": { + "parameters_hash": "1d997626a8076126a16ba4cdcf858406", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:26:45.842731", + "end_time_simulation": 59.97999954223633 + }, + "df4602c92bbd5b788625ea226da213fb": { + "parameters_hash": "c9bc82ff8253f11ee29d67e5be864987", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:26:48.193357", + "end_time_simulation": 59.97999954223633 + }, + "23ab2cac393646cb81c4aa4d6160ae73": { + "parameters_hash": "722536ccfb033a419c13d894e5ed44fb", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:26:50.951950", + "end_time_simulation": 59.97999954223633 + }, + "84213563f6bb59e7a227b801f9d6bc68": { + "parameters_hash": "5b25094f0095b03e02e73b618367e0d4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:26:53.819334", + "end_time_simulation": 59.97999954223633 + }, + "a04e965c08014d239d8b8a393e1a3cdd": { + "parameters_hash": "3572dd7ed11afe351f00c15cb16e62f1", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:26:56.374697", + "end_time_simulation": 59.97999954223633 + }, + "d60ab0ba03aa49b6a030173c4279f64b": { + "parameters_hash": "6b9ae2fd509f44e1e2fb55935871d864", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:26:58.915725", + "end_time_simulation": 59.97999954223633 + }, + "b45a4bc98b4a42fe91c2a752c7889172": { + "parameters_hash": "cb3bad6d21e2d77094916bd96ef4e273", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:37:33.924772", + "end_time_simulation": 59.980000000000004 + }, + "fab421c9d33553b7998f7615ccce8cd1": { + "parameters_hash": "afc3ae61cd927555b00e1dac180266f3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:37:37.835063", + "end_time_simulation": 59.980000000000004 + }, + "44f04c1b06cb4685b9878649fd7d6835": { + "parameters_hash": "ec10dd18a8601462bdc47a1ad069b3c5", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:37:41.870250", + "end_time_simulation": 59.980000000000004 + }, + "e9bca9d534e64f91ad16da863473aff1": { + "parameters_hash": "84ea90b63c3153ccfb9b943f6b60ccdf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:37:45.203240", + "end_time_simulation": 59.980000000000004 + }, + "74603e2ec58f5ea6b4cdd3e9c9335992": { + "parameters_hash": "4edad0b0aff0fbb5a23adedde7c13adb", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:37:48.768727", + "end_time_simulation": 59.980000000000004 + }, + "08c26ddcfac0570bb3c5e834931c5ced": { + "parameters_hash": "97647b66caf6287a45636750643d4561", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:37:51.465074", + "end_time_simulation": 59.980000000000004 + }, + "4640e354c5cc583fa840fdd67ddddee7": { + "parameters_hash": "a9ff143e93be0e44dd1283b240d703a8", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:37:54.252781", + "end_time_simulation": 59.980000000000004 + }, + "7b1ce6fe5bfd5ec58c6cede8164b7974": { + "parameters_hash": "a2893bf0eaac1ae0a1252b628404ae9c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:37:57.144538", + "end_time_simulation": 59.980000000000004 + }, + "779261e592034cf882881c2f70a20728": { + "parameters_hash": "f85f5510d85c3d8597a4b8a3fc177afa", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T00:10:01.915706", + "end_time_simulation": 59.97999954223633 + }, + "ca8e6694b045476e8d251f55f5445ab3": { + "parameters_hash": "df4762fc4837017b44d02f3d9712953f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T00:10:05.591663", + "end_time_simulation": 59.97999954223633 + }, + "6ccce0baa2684894915d00cd8c1eacc2": { + "parameters_hash": "848de3ce573c1b0385c820ecc3be015d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:08.418695", + "end_time_simulation": 59.97999954223633 + }, + "38acd0b1a27e480592074e16ec56f94a": { + "parameters_hash": "d3ca7aa3992c751eac12cf9c44ed2264", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T00:10:12.007502", + "end_time_simulation": 59.97999954223633 + }, + "8a41366cc9d350f4ab4e54502857065b": { + "parameters_hash": "0114b2d2eab8d98898302e3e756959fe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:14.877450", + "end_time_simulation": 59.97999954223633 + }, + "be6139ae18aa4266950a9e6450d6a507": { + "parameters_hash": "88cedf09f2af6ad32b588bbd5c4bd5ec", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-02T00:10:18.844984", + "end_time_simulation": 59.97999954223633 + }, + "1593c6f0b1c05209b09578523ebb4e12": { + "parameters_hash": "65d91d97f32a7561d82345734072c617", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:21.929294", + "end_time_simulation": 59.97999954223633 + }, + "e5d0ca629a2e49bcb76945a581cd1a98": { + "parameters_hash": "7affebd7a60baafa0c4cd5e84b9ef2ef", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-02T00:10:25.206546", + "end_time_simulation": 59.97999954223633 + }, + "c907ec38a25f47eca2d951607e1f5af4": { + "parameters_hash": "a6e873c162b7dfec2b96f4185e8ec289", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:28.166130", + "end_time_simulation": 59.97999954223633 + }, + "324f1712b21d47888e27b3d794abc71b": { + "parameters_hash": "635fe28f1f1a58be42225a9b61a1d56c", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-02T00:10:32.300700", + "end_time_simulation": 59.97999954223633 + }, + "92c8a46b2c5c4e62b19954b7786be203": { + "parameters_hash": "f49319c1100027acc349fbc5ea82b274", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:35.764016", + "end_time_simulation": 59.97999954223633 + }, + "f8dfcdfe548c4e16b2ba6c29f733f1fe": { + "parameters_hash": "4b12441c0fde8b2c0e763978a7cf41e2", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-02T00:10:39.133818", + "end_time_simulation": 59.97999954223633 + }, + "4d11f1790dd05e6eb137b05535652919": { + "parameters_hash": "9fc83986a0344c52fab0b66471478a07", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:42.339430", + "end_time_simulation": 59.97999954223633 + }, + "d107128850bd502784cdb511d7efa744": { + "parameters_hash": "ed5d8307e5f7968b39bcfd819b4d6d20", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-02T00:10:46.219547", + "end_time_simulation": 59.97999954223633 + }, + "29f08ce0f1745c77b89e559953928a0d": { + "parameters_hash": "e0a16a8fdadfcb99c92851419813402d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:49.314497", + "end_time_simulation": 59.97999954223633 + }, + "12e771aca3d051819794f29f4b99c004": { + "parameters_hash": "ddee832d6697585164a9f75940af6404", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-02T00:10:52.842229", + "end_time_simulation": 59.97999954223633 + }, + "b9d7c93d3d685d6e8883daaa537078f5": { + "parameters_hash": "65b0ad61fe61d65ceb388f91a6d1a02d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:10:55.446294", + "end_time_simulation": 59.97999954223633 + }, + "fab4f5f11b874ef58e909995690d57ec": { + "parameters_hash": "bad3ce5eb8272ff3ff26555f1ef53db7", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:00.317859", + "end_time_simulation": 59.97999954223633 + }, + "d55b284feeaf5437a227913fc3ce1f60": { + "parameters_hash": "d1bf4776a1fb98a93640bcc74cf381c5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:03.097664", + "end_time_simulation": 59.97999954223633 + }, + "da62c8d5d796459081533a11ff95d0c9": { + "parameters_hash": "b10e3c032bcc3535b1bed71322c70182", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T23:59:04.957485", + "end_time_simulation": 59.97999954223633 + }, + "949e5c008e4253cc99d7a184ce577778": { + "parameters_hash": "af6ca459a25640b0da0a49462b91fa87", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:07.711465", + "end_time_simulation": 59.97999954223633 + }, + "266ff8b228ea4c2bbc98b40b4b29ae1e": { + "parameters_hash": "1f67224e94945758df3fb47141ef3535", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T23:59:10.799439", + "end_time_simulation": 59.97999954223633 + }, + "1bb07ba9317c558780d60654c4d81916": { + "parameters_hash": "5138cf9622a398d0a026700b36f6b0dd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:13.481475", + "end_time_simulation": 59.97999954223633 + }, + "bc5ebe0657b44a568d2d3958f488be46": { + "parameters_hash": "bebec398e9f413561e686454e411f7e7", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T23:59:17.111762", + "end_time_simulation": 59.97999954223633 + }, + "e246a03f4a0c51b29595ce453ade330d": { + "parameters_hash": "07b446713ba79d6c8fd9fa40e66589bf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:20.069341", + "end_time_simulation": 59.97999954223633 + }, + "354dbe3c5b4b4ef29b2e2a1a0ae5c035": { + "parameters_hash": "243f64406b1770defea4dfd9855d4090", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T00:14:28.862247", + "end_time_simulation": 59.97999954223633 + }, + "8df479002605413a9947457a9631cc91": { + "parameters_hash": "a61357095af039980107a768863e06bc", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T00:14:33.364932", + "end_time_simulation": 59.97999954223633 + }, + "5864fec329f0567a8e31e0c924aa1f37": { + "parameters_hash": "f5a657b810f627086cfbbf24652dd980", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:14:36.470379", + "end_time_simulation": 59.97999954223633 + }, + "8f80071112174821a3595067c2bf1537": { + "parameters_hash": "0b8ddbea3b36ee43473118eb56d7ae9a", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-02T00:14:39.737557", + "end_time_simulation": 59.97999954223633 + }, + "54ea7b61faac5422a6babffc200b6be3": { + "parameters_hash": "7f7a7c6545f4d4125a72885b408afb01", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:14:42.265033", + "end_time_simulation": 59.97999954223633 + }, + "e95b6f3fd138444e8f024549b8f9f4bc": { + "parameters_hash": "301a02fdda96c9ba27a7976692c4123d", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-02T00:14:45.052102", + "end_time_simulation": 59.97999954223633 + }, + "64280e4d2e7c488ba2e87e8f2a166fa8": { + "parameters_hash": "0032190866fec5a20d54d76b94b788fc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:14:47.624223", + "end_time_simulation": 59.97999954223633 + }, + "f15e0273282540319a42a20f071b2766": { + "parameters_hash": "48688663173d6db8b6eb44af38e8c558", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-02T00:14:50.608632", + "end_time_simulation": 59.97999954223633 + }, + "e277deb5350248bdb634d6656bcf467b": { + "parameters_hash": "318d1d671306e7416c4f908a431affa8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:14:53.424856", + "end_time_simulation": 59.97999954223633 + }, + "051f28a4cee845c3a826eca9127ca028": { + "parameters_hash": "9a68648ef094a9fe0ec6da94f64eb617", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-02T00:14:56.930970", + "end_time_simulation": 59.97999954223633 + }, + "81e262226b2b4b9a8c35f16021170948": { + "parameters_hash": "4f8bd64d9a8023e5689da2b1703c7b99", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:14:59.396903", + "end_time_simulation": 59.97999954223633 + }, + "c67fd00576b04ac2bb3baf818c034ae4": { + "parameters_hash": "586e10e4d484c858801f077398ee7419", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-02T00:15:02.410883", + "end_time_simulation": 59.97999954223633 + }, + "21fd2e7cdbd944b2ab794e4d318f0008": { + "parameters_hash": "b7cae90a9be7ee524b8fe530be8ae1d0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:15:05.447501", + "end_time_simulation": 59.97999954223633 + }, + "76488608eab25e669b94aff029f43ea4": { + "parameters_hash": "d269d5adb15f8a199b1821bd7781f4f2", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-02T00:15:08.115386", + "end_time_simulation": 59.97999954223633 + }, + "d75a9624699a5b1ea775a80a3b19c148": { + "parameters_hash": "c5c0c91981bdd5da3f15b56c2c05487d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:15:10.585750", + "end_time_simulation": 59.97999954223633 + }, + "9d161ffe1cbe5b8eaa5719eaee5923c5": { + "parameters_hash": "0b54ecc0adc424a98743c424d0105281", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-02T00:15:14.458275", + "end_time_simulation": 59.97999954223633 + }, + "d75e1ceef1015432963f48a40a40aeb2": { + "parameters_hash": "3a5eb80613da2ee524dcd8d2c45a4c1b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T00:15:17.394136", + "end_time_simulation": 59.97999954223633 + }, + "88054c50f008401ba95e4bc68fb71ec8": { + "parameters_hash": "68f7e0f1a2ca1ca4dd95ae618ca06039", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:06.194327", + "end_time_simulation": 59.97999954223633 + }, + "3a4ac4546cfe510b96f90161b02ccd1e": { + "parameters_hash": "f29884bb6922eacaebd06078842b67f9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:09.135055", + "end_time_simulation": 59.97999954223633 + }, + "866e78a3aee946dfaab06893eac3d553": { + "parameters_hash": "f658b04327b8b0123de2350dc1f9fe23", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:12.094828", + "end_time_simulation": 59.97999954223633 + }, + "6ad9e14b7c635c82b160ff9b7e892c09": { + "parameters_hash": "e6159a5c278d48939b14977d4624af97", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:15.058731", + "end_time_simulation": 59.97999954223633 + }, + "105dab6037814698b8c5fb1b7cb0d358": { + "parameters_hash": "b8b3a9bced67d7217431b9939f2a9078", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T23:59:29.378686", + "end_time_simulation": 59.97999954223633 + }, + "bf3abf19cefc55c48be5f9ce10f27076": { + "parameters_hash": "d42ae37bbbb1def975a84551b5538337", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:31.888307", + "end_time_simulation": 59.97999954223633 + }, + "ed3abb63ca7e49dca5dbc6f41691f421": { + "parameters_hash": "9dbdbdef37ea51a4f6973c98511a5417", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T23:59:34.800397", + "end_time_simulation": 59.97999954223633 + }, + "a735122b9cd553ff9c807cdd11567cc2": { + "parameters_hash": "5af082cf4a467c61f3e17a21f44a73da", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:37.903750", + "end_time_simulation": 59.97999954223633 + }, + "7b2b2d6cce1c444d8fc6bc0be9ac42ee": { + "parameters_hash": "e76753a3799574c39b0e86afafa1b2ed", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T23:59:40.800846", + "end_time_simulation": 59.97999954223633 + }, + "75c9d0d6052e540d8ad0ca18b72eb129": { + "parameters_hash": "30374b4371c474affc90f4c3e23b3afc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:44.369961", + "end_time_simulation": 59.97999954223633 + }, + "934a4597164b46a7a2728e1ebf583b32": { + "parameters_hash": "a75236ebbaa2deb369c720d5a860c93f", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T19:42:36.358571", + "end_time_simulation": 59.97999954223633 + }, + "a71f474e41464187af46ccbd7222cae8": { + "parameters_hash": "e5c8469aedae7cbf8b54337f754e5ad4", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:42:39.258158", + "end_time_simulation": 59.97999954223633 + }, + "55edbdbe12ee4816b3c2f3e2ba9207da": { + "parameters_hash": "34ff3cb9956756188fe1f7936fa0c8fe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:42.314401", + "end_time_simulation": 59.97999954223633 + }, + "566239f610c74a3cb6dcf83e703d077a": { + "parameters_hash": "852b6231fdbff81c92eef1df73e79aea", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:42:45.088491", + "end_time_simulation": 59.97999954223633 + }, + "81b9c4bbd5fe59bfaaba8789ee286579": { + "parameters_hash": "04f481f18562fdbf89fddb0f6f169126", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:47.762250", + "end_time_simulation": 59.97999954223633 + }, + "319f9eb10a2b4290ad68cb6d0c975d24": { + "parameters_hash": "a84657bb7f98cb5817249962a46b40aa", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T19:42:50.791223", + "end_time_simulation": 59.97999954223633 + }, + "0390821eb97f52c6a29c77e7be7ade95": { + "parameters_hash": "10bfd6f9d3325c6af88a038221c352e8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:53.588687", + "end_time_simulation": 59.97999954223633 + }, + "3b00a79d1fff4b02a8566c05ad618150": { + "parameters_hash": "a291f1aabdec43557934be734a6a69ce", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T19:42:56.569371", + "end_time_simulation": 59.97999954223633 + }, + "c10106e83a974141834207ad9ee80094": { + "parameters_hash": "8609c72972d0ec8af8f23da2ec49b4bc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:42:59.550304", + "end_time_simulation": 59.97999954223633 + }, + "71479882752d490f9cd643c85968d4e0": { + "parameters_hash": "4261a4c62f01d6c0f3bd7fdcb738c87c", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T19:43:02.666749", + "end_time_simulation": 59.97999954223633 + }, + "b94fe00d78634d5c86f7c9bf4e505ef4": { + "parameters_hash": "3ae2ee0fcb02c915312abb4fe59490ac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:43:06.135569", + "end_time_simulation": 59.97999954223633 + }, + "182039f0b58a4ad590d08a1e9158e272": { + "parameters_hash": "6475cf2b37c7406569d88bb96ef63f33", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:43:09.678149", + "end_time_simulation": 59.97999954223633 + }, + "b91759b9d5014d77a4ce4003e191bf5e": { + "parameters_hash": "d60bbfecbb811a32eceb630546188771", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:43:12.731923", + "end_time_simulation": 59.97999954223633 + }, + "99da1cead2815b30be52dbaf541d5bd5": { + "parameters_hash": "6405eb398c4a074019ec87d0d1c02a09", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:43:15.589788", + "end_time_simulation": 59.97999954223633 + }, + "dc6c4fee3f4654aa9ec31a3537459467": { + "parameters_hash": "d2206f1ff91ad5b56760a7f8806028b1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:43:18.350468", + "end_time_simulation": 59.97999954223633 + }, + "cee872bc41b65da188f77efb937e7e47": { + "parameters_hash": "73d571c77e0ebf6b2e6c52382c4326eb", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:43:21.892374", + "end_time_simulation": 59.97999954223633 + }, + "35788fb01e495d88ba5af1cb7458da21": { + "parameters_hash": "9f56dacd5d80baa73985b38eb61caf4c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:43:24.714640", + "end_time_simulation": 59.97999954223633 + }, + "5838d4e3197541219e46a399693b4d50": { + "parameters_hash": "83cc529948817b70b62e5338cac2f443", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:23.918669", + "end_time_simulation": 59.97999954223633 + }, + "6a95f92a50635a65957b9fa6b0b7de6c": { + "parameters_hash": "6ce169791e22383118aca47ce5348935", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:26.246602", + "end_time_simulation": 59.97999954223633 + }, + "2310e88aa2a541b4a3a573100f28dcad": { + "parameters_hash": "7e6b89b42854d0c6b9dd6dabdb037bda", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T23:59:47.414406", + "end_time_simulation": 59.97999954223633 + }, + "986e1a4e7884536b8f62c6767cd050dc": { + "parameters_hash": "b3bb8c55718df198853cf99749c4287b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:50.430371", + "end_time_simulation": 59.97999954223633 + }, + "037081d7d6954010a36777c03d2d5447": { + "parameters_hash": "865bfa42c19aa388e9f05cbae734efcd", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T23:59:54.048084", + "end_time_simulation": 59.97999954223633 + }, + "c30ddff8c1995c89b3b0d5875dd675ce": { + "parameters_hash": "cb6c5dde063b569e2cd0f9bf1ece78f9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:59:57.308923", + "end_time_simulation": 59.97999954223633 + }, + "74b7a23252a94ac4aa7aa071d7b17aa5": { + "parameters_hash": "795785f4ba2c82450626bf9385860036", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:00:00.910535", + "end_time_simulation": 59.97999954223633 + }, + "aa5fc02266945f2ba9535b8f6ab7308b": { + "parameters_hash": "5c646926b7f31fb152b7703a434e4f91", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:00:03.967320", + "end_time_simulation": 59.97999954223633 + }, + "d67e94730c5146cead86b2fffb9109cc": { + "parameters_hash": "d102e7a91b00def881828bbab98af19f", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:00:07.100601", + "end_time_simulation": 59.97999954223633 + }, + "4d76951e2af952f8af98844321d473de": { + "parameters_hash": "7f5f77fcdac6d05c4fae5ee9021816de", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:00:10.044330", + "end_time_simulation": 59.97999954223633 + }, + "0272b7952029539d9bd40ce24d621991": { + "parameters_hash": "c61f31bc38e3b482158addeef8ba2e75", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T23:36:44.416744", + "end_time_simulation": 59.97999954223633 + }, + "72446adbf91542a2af4f3c79f40ab1db": { + "parameters_hash": "5e5d573eceb5b1417baa971705bc5e61", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:18:11.692287", + "end_time_simulation": 59.97999954223633 + }, + "16c5facd55b840f0bffbd9d0ba2db717": { + "parameters_hash": "8510d6d9db9fb5e909a2461b10732544", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:15.307125", + "end_time_simulation": 59.97999954223633 + }, + "95c5547b349a5e5f90f7b4eac42a205a": { + "parameters_hash": "3231fb3ca0b491b91171e8c96dcebd8c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:18:19.185058", + "end_time_simulation": 59.97999954223633 + }, + "0c25b554248c5eef80ad123763baa0ee": { + "parameters_hash": "14afd3924f27c979630621113a23841c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:21.961960", + "end_time_simulation": 59.97999954223633 + }, + "9730eb23e2d751f6886c4a0d58d7e5e8": { + "parameters_hash": "f67103c1317c62434ff0f37d718c5089", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:18:25.446845", + "end_time_simulation": 59.97999954223633 + }, + "aaeb12dc17c354b6a8e3a84379ee3402": { + "parameters_hash": "6a39189f1fef9eeebc4adb90bd9e4dcd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:28.070257", + "end_time_simulation": 59.97999954223633 + }, + "c965fbd0f3674a23bd71a066a659f927": { + "parameters_hash": "b96ebd0586244d7b9c63b16866059330", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:18:30.813019", + "end_time_simulation": 59.97999954223633 + }, + "8cff48e34ad9496ba4d5921130f28d04": { + "parameters_hash": "30c6bfb94046de41423b52dfaebbeaac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:33.638036", + "end_time_simulation": 59.97999954223633 + }, + "148cb65464985cf09acc84595e7a7ae3": { + "parameters_hash": "a97fdbdc4907bef0fa4d576e91eafdc0", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:18:36.259503", + "end_time_simulation": 59.97999954223633 + }, + "ac166f240b50588cb94fc59bbe10613c": { + "parameters_hash": "dba83b992cf0f8f3acac0ae6639ee1e2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:39.132278", + "end_time_simulation": 59.97999954223633 + }, + "2f760a132640578b96960b7cee8ae277": { + "parameters_hash": "df46b10959e0653936736c9ad43bbfe1", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:18:41.903981", + "end_time_simulation": 59.97999954223633 + }, + "95759e5d11cb5034946bb32e20d7ba57": { + "parameters_hash": "3fcd60b63a1340692e530a054a44cc22", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:44.926017", + "end_time_simulation": 59.97999954223633 + }, + "f4db731dca6e563a8333b5c200aefd3f": { + "parameters_hash": "7a1edd8435eeab8623d3476af2d383fc", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:18:47.763591", + "end_time_simulation": 59.97999954223633 + }, + "a827069d6ff453a4a7e9532adaa2e6b2": { + "parameters_hash": "5402f035c30c482098e94fa3c3bb75ac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:50.756184", + "end_time_simulation": 59.97999954223633 + }, + "8399e56763da5bc59ef9f9c4b652eeee": { + "parameters_hash": "5b3926fe15045159c0fef5628c86baee", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:18:54.065674", + "end_time_simulation": 59.97999954223633 + }, + "b316cdf32a6b51c98365e0b0f8462b68": { + "parameters_hash": "d9b5ff47f0fb48172b9acb936926091d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:18:57.025694", + "end_time_simulation": 59.97999954223633 + }, + "22de2bca0b20512e9945f056ecb28478": { + "parameters_hash": "6eef8ce1ce995571ec31cdcf0852a310", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-06T00:31:55.128525", + "end_time_simulation": 59.97999954223633 + }, + "7b009af4d3ce4310ad79cf5329d76c13": { + "parameters_hash": "b9e95dec433a463e36efd96fa46ccd19", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:31:59.761518", + "end_time_simulation": 59.97999954223633 + }, + "b68b0db6872d4e12a298b33951a65e51": { + "parameters_hash": "4a0c54665c5fb421c4eaa01b9b369d9f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:02.940261", + "end_time_simulation": 59.97999954223633 + }, + "668aa669116c5beead5db1df59c962c1": { + "parameters_hash": "223999cabf58fa0b070be58d6e58ecf9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:32:05.798367", + "end_time_simulation": 59.97999954223633 + }, + "b8b13f06afae5f85b3d86585c46aae28": { + "parameters_hash": "d8bd0c540d66aa677ab1cc8da4b55205", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:08.484543", + "end_time_simulation": 59.97999954223633 + }, + "365154f304255455930bc8495d8ad916": { + "parameters_hash": "e790e8badcce8d9f1c3aba89508939c0", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:32:11.077317", + "end_time_simulation": 59.97999954223633 + }, + "f619adc8837a5c10936ce76a3f0bb464": { + "parameters_hash": "f4d7b4c616945f51339e13e1144ebd71", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:13.671253", + "end_time_simulation": 59.97999954223633 + }, + "6dfb61e455f44fb3a3ebaae7dfcb1604": { + "parameters_hash": "61eac9424dc39ac64c069ac092acaa6d", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:32:16.686569", + "end_time_simulation": 59.97999954223633 + }, + "d133bdec5871452ab7cd7a84a2423b48": { + "parameters_hash": "cee02a474391c015ff40c83401be5405", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:19.512181", + "end_time_simulation": 59.97999954223633 + }, + "febc5f7c2e2755d9a9fe80e01ccceda1": { + "parameters_hash": "f100a327ea6a32dfa2ba9ddfd6f7e352", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:32:22.175360", + "end_time_simulation": 59.97999954223633 + }, + "03c2924a8e8b541599d6d3c977b3de7c": { + "parameters_hash": "ae872918a508913b64c653c93511860f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:24.973081", + "end_time_simulation": 59.97999954223633 + }, + "a28790cad25c4135beed90ab69e85f70": { + "parameters_hash": "7a170c288a692177ac340b544278fef2", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:32:27.634403", + "end_time_simulation": 59.97999954223633 + }, + "2f798da8cfdc4896ab11b7c00f7636e1": { + "parameters_hash": "721cf4bdee245714b4dee46a2ec0f5ea", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:30.051294", + "end_time_simulation": 59.97999954223633 + }, + "0cbfd492d485523abab3cefd5373e9c8": { + "parameters_hash": "58c45fee76d499fce845e997f423eac8", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:32:33.061982", + "end_time_simulation": 59.97999954223633 + }, + "29e4556480e758f6b08785b704e5d7cc": { + "parameters_hash": "9037a8208078cf1006fac26915adaa70", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:35.693615", + "end_time_simulation": 59.97999954223633 + }, + "571418865cc7562088e7d1f7d21d96d1": { + "parameters_hash": "5676be7605ba578ab659b9350a657a2e", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:32:38.276474", + "end_time_simulation": 59.97999954223633 + }, + "0c656a33f151503eac50fbbd6e3b3907": { + "parameters_hash": "30affe7d755722d3b7bab2c4fe13ec74", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:40.687186", + "end_time_simulation": 59.97999954223633 + }, + "b4c09e6da7225bc797c0e8e86c5d3cd7": { + "parameters_hash": "0b6a9a8f26f864d40c94d266472b91a5", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:32:43.423910", + "end_time_simulation": 59.97999954223633 + }, + "0fd1e8e21998582a9669dfb4ebb8ac19": { + "parameters_hash": "442bf6c7a87fe7cb0957e5e79eb78cfb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:45.604653", + "end_time_simulation": 59.97999954223633 + }, + "0272541246be51b9874fb7f7ff6c1955": { + "parameters_hash": "d31a7cfcc28140ee73e0b08c2645296d", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-06T00:32:48.060531", + "end_time_simulation": 59.97999954223633 + }, + "6fe33d3ed9c8435fb9fa5d1fa37b3fb4": { + "parameters_hash": "4c5fce5978f9f2a2634b7887c72dc74c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:32:50.853583", + "end_time_simulation": 59.97999954223633 + }, + "41fba608fa404eb1acd3a08b128c8742": { + "parameters_hash": "d6022d4a9485cb0b223b94a8e06eeb11", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:53.298873", + "end_time_simulation": 59.97999954223633 + }, + "4f32b93768d25e5a8477fadb466e1fd3": { + "parameters_hash": "2510f28dd65ded28f9e7069546587112", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:32:55.980547", + "end_time_simulation": 59.97999954223633 + }, + "76de0be5173e5076b0f557b1a555b52f": { + "parameters_hash": "63a9a35da9c8c072e32c3fb6e9cd5d0a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:32:58.555978", + "end_time_simulation": 59.97999954223633 + }, + "0c5e4d9eed4b57618d9987379492253a": { + "parameters_hash": "7ea42b200bf05b634f026d5685a986b2", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:33:01.191556", + "end_time_simulation": 59.97999954223633 + }, + "0f11efda45a55aa7968b3bd282106967": { + "parameters_hash": "20c3760a26b46ee4c4124f6849c4b534", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:04.049442", + "end_time_simulation": 59.97999954223633 + }, + "cf818cc6219a4dad9783ccfa3afc9686": { + "parameters_hash": "3f103bb892b3af5a9b01add6b9e3d44f", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:33:06.580928", + "end_time_simulation": 59.97999954223633 + }, + "e38756e03f68512d9b724a4ea10756b7": { + "parameters_hash": "7d1a162afe82d0458de9a640034a2451", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:08.934222", + "end_time_simulation": 59.97999954223633 + }, + "0d1518c39bf74069810673be6f916975": { + "parameters_hash": "13c754a7a43b3e4361810dff77e4c186", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:33:11.851439", + "end_time_simulation": 59.97999954223633 + }, + "2ef5aa5aceb64a36a73873bbbd573056": { + "parameters_hash": "2b8161e0bbd25a02ab66ca60cf917993", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:14.445215", + "end_time_simulation": 59.97999954223633 + }, + "4ad1f6766dd2542e93768ffc77bbbbb9": { + "parameters_hash": "3768818365a06691586906a1e3c0e049", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:33:17.233357", + "end_time_simulation": 59.97999954223633 + }, + "6a38bccfc82c56a7b32de22d10216af5": { + "parameters_hash": "ba7c61b5a701a1588a4b5e626edfa32e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:20.310326", + "end_time_simulation": 59.97999954223633 + }, + "71392794c60043aa94c561f678b51c96": { + "parameters_hash": "50d488f9d46f854762b432a3e620b3ca", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:33:23.100271", + "end_time_simulation": 59.97999954223633 + }, + "e41d021c70754bc59dc512dae58bcc45": { + "parameters_hash": "469923bd51bc28ff8fd1de47683254d8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:25.654074", + "end_time_simulation": 59.97999954223633 + }, + "c940dcc023d15ff49ba22c06b11dfe56": { + "parameters_hash": "15680192eae74028ffeb540d5e606ca4", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-06T00:33:28.258391", + "end_time_simulation": 59.97999954223633 + }, + "790f3a4580d05107855f7a4a05b98502": { + "parameters_hash": "f74ae07af9b4b6c1edd4f6abc7319624", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:30.826541", + "end_time_simulation": 59.97999954223633 + }, + "b91c88b11ade5300bd33317f8db6200a": { + "parameters_hash": "2f2ae2f69fe5fb61b74c86eb8488a1ed", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:33:33.674726", + "end_time_simulation": 59.97999954223633 + }, + "854b15cfe6355079854665f7afd485fd": { + "parameters_hash": "13170c83ca4304fd8ea2b6ec599fdea0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:36.119858", + "end_time_simulation": 59.97999954223633 + }, + "05e25405ff135208b57eb971eade3c72": { + "parameters_hash": "99ef7003fa1b91d69a89fe02ab91abfd", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:33:39.135919", + "end_time_simulation": 59.97999954223633 + }, + "be7e51015fce58769b9449e9a305478b": { + "parameters_hash": "11c294d003b3b6f0975ac62ff1cb6786", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:33:41.645091", + "end_time_simulation": 59.97999954223633 + }, + "b731aec1c1145504a518ce5b1951c95f": { + "parameters_hash": "19bbd66bbf5154d26016bc67519648a4", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T19:48:13.250406", + "end_time_simulation": 59.97999954223633 + }, + "052441f7276d4a23988fd4b69bc56d9d": { + "parameters_hash": "279b7eb932a93fe67823001dff98f068", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:48:16.039436", + "end_time_simulation": 59.97999954223633 + }, + "920b2bbb7e8746698c43f537bbfbd4d1": { + "parameters_hash": "14edbf6fdc59f3c78320f7e537fd1488", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:18.375343", + "end_time_simulation": 59.97999954223633 + }, + "5300e206a01255ac86bbc4b0251cc195": { + "parameters_hash": "89f59dd043a97719a16a6f28b31a3f03", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:48:21.763758", + "end_time_simulation": 59.97999954223633 + }, + "68ed490ffd8c58feb682be08fb4d5156": { + "parameters_hash": "b638d48688904f5227169de65142e817", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:25.132877", + "end_time_simulation": 59.97999954223633 + }, + "4bca4a7829715d58ad63c7da524e7341": { + "parameters_hash": "7fcd1ae9ed2fbbe358326a0bcad4fef6", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T19:48:28.371082", + "end_time_simulation": 59.97999954223633 + }, + "a6e109e2b5b754cda74f10fb3b114251": { + "parameters_hash": "a7fa372a90e1d07ec5ff42df57bd2380", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:31.240972", + "end_time_simulation": 59.97999954223633 + }, + "58c2ee5373605713b6cc06cb3e346ef9": { + "parameters_hash": "3038575e253afb6ea88bbd014ea46f19", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T19:48:34.656813", + "end_time_simulation": 59.97999954223633 + }, + "b0d8f930c4855e5eb252af549e49a5fd": { + "parameters_hash": "99d2c7a4e422cc133948dea90e462b53", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:37.927443", + "end_time_simulation": 59.97999954223633 + }, + "f667af886e0f422587b2dffeaf23fa73": { + "parameters_hash": "a6e07af4e26bb2ce561d1a7fab087a9a", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:48:41.170426", + "end_time_simulation": 59.97999954223633 + }, + "7e889ca35e5546788f3e6a84b9ecc0cf": { + "parameters_hash": "47f012ca41743c73f8cf5b28b832a04c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:44.049418", + "end_time_simulation": 59.97999954223633 + }, + "da2576824c005190bd1a17d25d7790dc": { + "parameters_hash": "7573ba2d20d38196f3721b7ac9deea84", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:48:46.916356", + "end_time_simulation": 59.97999954223633 + }, + "3355f00032255d6ab57cf77cc161fcbb": { + "parameters_hash": "c2ac6fc391c87c092c8e00e225ac8105", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:49.723545", + "end_time_simulation": 59.97999954223633 + }, + "ef4ceccd93225972aff4165c0d622f57": { + "parameters_hash": "094a7b751a03b2f19e89de091ee8a08d", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:48:52.854117", + "end_time_simulation": 59.97999954223633 + }, + "81647e9092fb5db083ee74fcf765a0b5": { + "parameters_hash": "cb07ae6f13c205cdd357e64f03d4762f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:48:55.600849", + "end_time_simulation": 59.97999954223633 + }, + "ff0d3964c5b954e6b77d9feab0f76e28": { + "parameters_hash": "e4188ff3ebfa16455982cbc5e2d37654", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:48:58.186340", + "end_time_simulation": 59.97999954223633 + }, + "0cf6ab9463f15ab7b30b323e2b46c63c": { + "parameters_hash": "7654bc0eae035f2993ff6e1891843eca", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:49:01.185298", + "end_time_simulation": 59.97999954223633 + }, + "6d6e5f8faa79553cb3b6e6f4c31d4de2": { + "parameters_hash": "36aa6bf3374c05e1d6378adca2023921", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T23:40:25.479027", + "end_time_simulation": 59.97999954223633 + }, + "c87fd1dc6d72419c83c36f9ba974fe05": { + "parameters_hash": "bd58e6ac233b23bf7f304c8c20ea6983", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T23:40:28.632770", + "end_time_simulation": 59.97999954223633 + }, + "44d0f0466a4a47e3a31ef87495b2d948": { + "parameters_hash": "72d9150e94753273f0c64906c8eb1a3d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:40:31.392451", + "end_time_simulation": 59.97999954223633 + }, + "70a3c8d105bb56759d43416418850b93": { + "parameters_hash": "2bc33556e5dc3b5d0ea79eb58bd1a941", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T23:40:34.469366", + "end_time_simulation": 59.97999954223633 + }, + "cceb21884d4e5f3d99b002294aca6a00": { + "parameters_hash": "fdea7434421e990f739a785e2219d208", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:40:37.141481", + "end_time_simulation": 59.97999954223633 + }, + "7a4e81fc68c55be88ab006073ee3c4f9": { + "parameters_hash": "5efa6a31bf36c3bf68ba0cc2a95c9534", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T23:40:40.558855", + "end_time_simulation": 59.97999954223633 + }, + "d605e2dc7ae75fbda6c4027c007bdee9": { + "parameters_hash": "2f54033affdbd65934f468ad918e7799", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:40:43.053500", + "end_time_simulation": 59.97999954223633 + }, + "b075618b4aa45099949b5a072cc9be7e": { + "parameters_hash": "97c2148077d20099eab17a24b0879dcd", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T23:40:47.067861", + "end_time_simulation": 59.97999954223633 + }, + "0923acf7779c5f5f87776e48dd34e5c3": { + "parameters_hash": "9f20992bea12629cd49eb99a0b162a33", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:40:50.248815", + "end_time_simulation": 59.97999954223633 + }, + "cc080755498055de93cebb72440a86c6": { + "parameters_hash": "8e401738fdbc28eb83c176ceb4cad975", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T23:40:53.562698", + "end_time_simulation": 59.97999954223633 + }, + "fc53ad15def95e5f967791d40ec2c6e9": { + "parameters_hash": "ad70feee8d494fb93b4396d93293be6a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:40:56.554593", + "end_time_simulation": 59.97999954223633 + }, + "7658e84d0dc755759d66adc3adf890c0": { + "parameters_hash": "f6058b78bd2f66fe198c0e4ad88085c5", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T23:40:59.682460", + "end_time_simulation": 59.97999954223633 + }, + "46949383f8695d709e150c1186184d90": { + "parameters_hash": "fa91bf991027166e9034083279a5a61e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:41:02.180918", + "end_time_simulation": 59.97999954223633 + }, + "9584da5231b4512188b5089c9a0dbc05": { + "parameters_hash": "0c472fe8fc6a82d3c8973b210920a1f5", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T23:41:05.320752", + "end_time_simulation": 59.97999954223633 + }, + "ed52695da75c542f9cf8ff67541e4a6a": { + "parameters_hash": "f04d82b8cfbad43e2d1ac9e0467a6b8a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:41:08.279463", + "end_time_simulation": 59.97999954223633 + }, + "d31bac7398034409ad2a8d05cf5ea3f8": { + "parameters_hash": "1e71612c7615089a03357efd6bf8ec9f", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:29.169867", + "end_time_simulation": 59.97999954223633 + }, + "f86e334a10985f67b1d0d6b9593c5f3c": { + "parameters_hash": "7264a639da947dd3c2c34d08f954f54e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:32.051218", + "end_time_simulation": 59.97999954223633 + }, + "a3d4d1a02f6f4d74a743b6f0212ec691": { + "parameters_hash": "a362aefef1a80b9fd63926fbb535ce10", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:35.411229", + "end_time_simulation": 59.97999954223633 + }, + "241c4c22ace65ee68c67bb063cb240c6": { + "parameters_hash": "76967fd02a770b3c7abfbc393fff5b8c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:38.525363", + "end_time_simulation": 59.97999954223633 + }, + "a0399f67cb5f4165a4dc00fa72104bb6": { + "parameters_hash": "a4e333a6de8f677270489817573748c9", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:00:25.189962", + "end_time_simulation": 59.97999954223633 + }, + "c35d8b6529ed5fc5b5a5866a0366156b": { + "parameters_hash": "2e9b7f1f379166a4bbb74728fac70292", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:00:28.510607", + "end_time_simulation": 59.97999954223633 + }, + "7bf12049ad514bca834e7b889c7cae6c": { + "parameters_hash": "ce4e26f1a48a9d245c4917935c57780f", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:00:31.735344", + "end_time_simulation": 59.97999954223633 + }, + "9c98eb0e013d5699b072fb5fcbaf3370": { + "parameters_hash": "1fe2037e6b4c2d9272c803a44d0d963b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:00:34.595808", + "end_time_simulation": 59.97999954223633 + }, + "1032bb73c0b55d1ba0c4685bb20f08f6": { + "parameters_hash": "01700931d35a6e95b56e0c2f17f9d088", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T15:09:10.516364", + "end_time_simulation": 59.97999954223633 + }, + "47de6546ee9b403881d7d758f0e0c970": { + "parameters_hash": "dd38617fc2f3e37fc4df4e43115a6955", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T15:09:15.936515", + "end_time_simulation": 59.97999954223633 + }, + "ef26a3713b0a426988fe1216e0067390": { + "parameters_hash": "9a4735ac79afd994a5c5e82f1f8349df", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:09:19.556494", + "end_time_simulation": 59.97999954223633 + }, + "7ecb534b8174567e92f87b88ab6ce793": { + "parameters_hash": "5ba0cad274028533bf91d759047c5a2d", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T15:09:23.313182", + "end_time_simulation": 59.97999954223633 + }, + "fefc73b45ddb5e6e85b32a5122c839d4": { + "parameters_hash": "c171b1b17b21778bbb2820f979943dff", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:09:26.714950", + "end_time_simulation": 59.97999954223633 + }, + "600a4df966dd50a6839a6d4a25267f60": { + "parameters_hash": "5fa2316d64af685fdb8d2744c83cad01", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T15:09:30.884273", + "end_time_simulation": 59.97999954223633 + }, + "56b8a2186c8c5d5e95be288ed368dea5": { + "parameters_hash": "84e5232dfa76e1376c330dbaa8fe201b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:09:34.535050", + "end_time_simulation": 59.97999954223633 + }, + "77bfe2ff37785558aa3411cc4f02bb4e": { + "parameters_hash": "c0e1d44ba96ab5271a437584edddb343", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T15:09:38.917857", + "end_time_simulation": 59.97999954223633 + }, + "63a9bcb84d305be098bc6bbb32b86c45": { + "parameters_hash": "5e78d7b8406d5d60c486986cc853f4e6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:09:42.253054", + "end_time_simulation": 59.97999954223633 + }, + "2027cfe28b64559995e4951b4f15d04c": { + "parameters_hash": "c73a30ea202fa5d5947b2e4e3d9ab11a", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-04T15:09:45.081583", + "end_time_simulation": 59.97999954223633 + }, + "6a677650bf6b51a59fb191ca25f2b5de": { + "parameters_hash": "b8616de8cf556adf9fa0085a9b45140c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:09:47.416960", + "end_time_simulation": 59.97999954223633 + }, + "7512635bb6e4583dbfde534576dea750": { + "parameters_hash": "f7bf92f15c30cfa10f94b01c73417a36", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T15:09:50.073268", + "end_time_simulation": 59.97999954223633 + }, + "3a44a2165aca52ee9daf9177477bbc71": { + "parameters_hash": "11c22dae9713baf9aef2ce4e9eb8cf65", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:09:52.699644", + "end_time_simulation": 59.97999954223633 + }, + "3a901bfee163599cb81f0751418e5ece": { + "parameters_hash": "537013e33ad9f4cafe86fc16f07c6bad", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T15:09:55.499039", + "end_time_simulation": 59.97999954223633 + }, + "c4d5f1b606065310ad33d26bd76bfeb0": { + "parameters_hash": "17d93794827780ac5cff869a105cfec6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:09:58.483980", + "end_time_simulation": 59.97999954223633 + }, + "105de8c7a17f4728931344dd36a28313": { + "parameters_hash": "1e55b3e782900756be811ce23aa9c293", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:41.885774", + "end_time_simulation": 59.97999954223633 + }, + "7b158201668351e0aeab4873c1edba76": { + "parameters_hash": "95d5b90d376e4eae0f56434c346c309a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:44.839880", + "end_time_simulation": 59.97999954223633 + }, + "c5b764da8f9442ebbce3b0859ed481a1": { + "parameters_hash": "ed4fb51fbbe9085bcc26ab5ec9f832a4", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:48.303964", + "end_time_simulation": 59.97999954223633 + }, + "05b2e8a8e9205b03924e50f0a6e98636": { + "parameters_hash": "8d021cbaab1925796599a3ee26de5bd0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:51.544333", + "end_time_simulation": 59.97999954223633 + }, + "7976849e5e764db4a0c080c98b14a478": { + "parameters_hash": "03ac64f6410c63b79be31bf7d1dacb06", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:00:49.897621", + "end_time_simulation": 59.97999954223633 + }, + "d344f859b10c5a72a39e8e309baaaaf9": { + "parameters_hash": "f80cc37b8395d1e5c571cfdb2c6e2b98", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:00:52.260978", + "end_time_simulation": 59.97999954223633 + }, + "855c0b4b25c947ddada661a7b2138f42": { + "parameters_hash": "59667920e886f04b97b7074efbe2121b", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:00:55.455488", + "end_time_simulation": 59.97999954223633 + }, + "ae4f7939de375836963dfbe3344529d3": { + "parameters_hash": "32b9c7e63e6fa22951f43991706ad365", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:00:58.469679", + "end_time_simulation": 59.97999954223633 + }, + "cd37b961b56951ce9d257f05362cdba6": { + "parameters_hash": "3029237a2e32748b2f4974409ea0afac", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T23:57:49.172268", + "end_time_simulation": 59.97999954223633 + }, + "74959bcdb9ad476f8871f21dc2ca5ba7": { + "parameters_hash": "8c955707cd3ab5af378e24fd60d214db", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T23:57:52.166280", + "end_time_simulation": 59.97999954223633 + }, + "e36e8b0d0d2d4cdb882f7c7eced19686": { + "parameters_hash": "4e1303b3f1df29a51fc160ced777b1ca", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:57:55.840162", + "end_time_simulation": 59.97999954223633 + }, + "0f637ad0c9985dbea90f9108824758a4": { + "parameters_hash": "1c0f37590ba2ceb08632f15a0441ee9a", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T23:57:59.130394", + "end_time_simulation": 59.97999954223633 + }, + "71b08374cbb25e0e83eab1bb83914fde": { + "parameters_hash": "1df69022ecb3f1b3e9a452328e68cad9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:58:02.391846", + "end_time_simulation": 59.97999954223633 + }, + "00e0a7f08ad750bea8c9d4a0a9df01bd": { + "parameters_hash": "f24135fbcb90cdd0291615589c7250fb", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T23:58:06.319594", + "end_time_simulation": 59.97999954223633 + }, + "878f9cc3940f557a8d7b3ef438bb4836": { + "parameters_hash": "ef66b11013d546c84ea00001efb3ac8f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:58:09.409900", + "end_time_simulation": 59.97999954223633 + }, + "1cb51571767c49d589b7640fd12dd132": { + "parameters_hash": "0af26206860915c1cd4b2a5f99210895", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T23:58:12.551771", + "end_time_simulation": 59.97999954223633 + }, + "870d8308a81c4dbc8db580047908836e": { + "parameters_hash": "977eebe7588d2d1e929fe12b821fec6a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:58:15.781069", + "end_time_simulation": 59.97999954223633 + }, + "d756c24fa5ca5af99f8a329d08e9456d": { + "parameters_hash": "3b3c8db0f203a214fdec55f2decc4f21", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T23:58:18.885925", + "end_time_simulation": 59.97999954223633 + }, + "a56d01cba1445d08ae8ad71453e1808a": { + "parameters_hash": "3aec8a1c196dee60d5893389dc30ee16", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:58:21.623155", + "end_time_simulation": 59.97999954223633 + }, + "65d4236d2ea14025b57278cb1b684df9": { + "parameters_hash": "6e5c1a5baa960103d0cca47fde9a9376", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T23:58:25.429518", + "end_time_simulation": 59.97999954223633 + }, + "50819bf381054abcb076204bc5f578b3": { + "parameters_hash": "fe77ffa517b26345f26aff7c058cde42", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:58:28.090600", + "end_time_simulation": 59.97999954223633 + }, + "73570e698c3c5595bb41dfb6c2ce3eb4": { + "parameters_hash": "aa290f3f6df591993d10116f8903a44e", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T22:18:48.162641", + "end_time_simulation": 59.97999954223633 + }, + "d0e82b242e8556c2bd4e1953f573956c": { + "parameters_hash": "a710fdd7c2ec6f8d14b5275fc05a8ae4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T22:18:50.813939", + "end_time_simulation": 59.97999954223633 + }, + "4640f134cb07578b9c5ce39acba95c8a": { + "parameters_hash": "522f8be6e590e22e18d360d654bb3aaa", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T23:58:31.202292", + "end_time_simulation": 59.97999954223633 + }, + "298ab37f6f8654049b3eecfe6d32d430": { + "parameters_hash": "4782051c5a0d796e73f25c3169a8e74a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:58:34.514846", + "end_time_simulation": 59.97999954223633 + }, + "19601357c3635302941fdc4f32321ab3": { + "parameters_hash": "509aa094502d2ca898ad443fef4a1345", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T23:58:37.738284", + "end_time_simulation": 59.97999954223633 + }, + "84799f45f20354a1b5e58a10414ee227": { + "parameters_hash": "75f1b6ca8a5cbe02825f84dba1a1b159", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:58:40.476924", + "end_time_simulation": 59.97999954223633 + }, + "a62b115d2796558ea6e26baf7f750b10": { + "parameters_hash": "7c6c73dd0ed3b43242fc53053862f219", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:52:35.641019", + "end_time_simulation": 59.97999954223633 + }, + "43e4aa8477c9421f8639250fc8ee535c": { + "parameters_hash": "cf9a44a25884f35b4246cc5773fd79ba", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:52:37.915249", + "end_time_simulation": 59.97999954223633 + }, + "fd641c7dbb694eb5b20b90c73a8a6b8c": { + "parameters_hash": "f874e403e8214c8b47f2ece91fa45156", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:52:40.074060", + "end_time_simulation": 59.97999954223633 + }, + "26c8949681765e7bb18863bd2fc39172": { + "parameters_hash": "21c9c09af980f6632ddfb36bd2b45813", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:52:42.283305", + "end_time_simulation": 59.97999954223633 + }, + "712a2f16c0675cf79899156788d7bc8d": { + "parameters_hash": "81b27cf386578ed060e80c091e7492a7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:52:44.274970", + "end_time_simulation": 59.97999954223633 + }, + "4e79aa9be3965a2ebb71f6d14ee04803": { + "parameters_hash": "561f0a56bcd5a75346fa5f3bfb5bec9d", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:52:46.396671", + "end_time_simulation": 59.97999954223633 + }, + "b3617ba1646d599bb2a100df8605cbcc": { + "parameters_hash": "5a26c22c74aa58f5445d42d7d661064c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:52:48.530940", + "end_time_simulation": 59.97999954223633 + }, + "c448dedaa70d5010b110245b7a687b5e": { + "parameters_hash": "452fb0c71f58e05f4bcd49530b507e32", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:52:54.992075", + "end_time_simulation": 59.97999954223633 + }, + "ff62314bb9df5ff08e3c94073dff06d5": { + "parameters_hash": "d1cb55c33350826e87212d2ba663e42f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:52:57.057028", + "end_time_simulation": 59.97999954223633 + }, + "42a2aa5119bd5b02aaa4a0e7a36246ae": { + "parameters_hash": "7e79c1532485e91a020392a5ce9202d0", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:52:59.179763", + "end_time_simulation": 59.97999954223633 + }, + "1a9edec3de635613b6e342aba59359db": { + "parameters_hash": "8cb3fba044952b5d6995384e622c44f3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:01.268490", + "end_time_simulation": 59.97999954223633 + }, + "a28fe88dd4585192850df4382159e5f8": { + "parameters_hash": "773ec891dd55f6aa9c069b5b525e4d59", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:53:03.553642", + "end_time_simulation": 59.97999954223633 + }, + "7684489a70fd5416bd1d6713cda57114": { + "parameters_hash": "313d5b8a79610e9abe8a970df95aa92a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:05.759042", + "end_time_simulation": 59.97999954223633 + }, + "94bf7ebf7a7b5df2ad630aa707610ecf": { + "parameters_hash": "27ca02862d855c85940962b3c6c099eb", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:53:08.018767", + "end_time_simulation": 59.97999954223633 + }, + "05c59769f6a958e09c70bfb091ba600a": { + "parameters_hash": "430f0bfdf0a745afed8fc3a4a4c3fd4a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:10.242552", + "end_time_simulation": 59.97999954223633 + }, + "ef3b6622f0be56338b9e4983386ea1d8": { + "parameters_hash": "64f91f13b1ab6f2b4d4973a186a5b7d0", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:53:12.334137", + "end_time_simulation": 59.97999954223633 + }, + "89871024495e4120b7de54369d852e26": { + "parameters_hash": "e393eab27e2d8eacb9d847ccf9b949d9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:53:14.604495", + "end_time_simulation": 59.97999954223633 + }, + "779ba224fabd428ea2c229339040dbe3": { + "parameters_hash": "c5371fddf49ce1f65beed191aacb5c59", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:16.670894", + "end_time_simulation": 59.97999954223633 + }, + "48293c790e0555dc809d2c46013e182e": { + "parameters_hash": "32b376caaee5176f3ca07b0404b2a8d8", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:53:18.910177", + "end_time_simulation": 59.97999954223633 + }, + "453fa34cb4ed5511a99c60a0911ed940": { + "parameters_hash": "4036dd06fa67a75487b5b78384f41f04", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:20.882014", + "end_time_simulation": 59.97999954223633 + }, + "bc7e217d7407451bb5c5fef16c661a78": { + "parameters_hash": "95ccc2baaca6c601a12a55c4a3de6f75", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:53:22.977532", + "end_time_simulation": 59.97999954223633 + }, + "89cf2f986032468b9940a99c802ca290": { + "parameters_hash": "810458a61661b62ec570def7489e0f50", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:24.959834", + "end_time_simulation": 59.97999954223633 + }, + "90896c4f728d53699700f6a9b97e63fb": { + "parameters_hash": "845f6078d939683b68cf9849b226a969", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:53:27.244695", + "end_time_simulation": 59.97999954223633 + }, + "85d2f37ffbae53728b86bb9c39325258": { + "parameters_hash": "dd28df7451cd5f955789963b0fd40413", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:29.218207", + "end_time_simulation": 59.97999954223633 + }, + "c3a9f28cb88d41aa9b42b02c003ad0f8": { + "parameters_hash": "e492a80fb679e985aa68cbcf4f912349", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:53:31.582035", + "end_time_simulation": 59.97999954223633 + }, + "ff27ed0616b242eb8f8b18146170552b": { + "parameters_hash": "21fa636ea5ee6eff55e55de5706e9fa9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:33.512257", + "end_time_simulation": 59.97999954223633 + }, + "995d020dee665552b68114218e00042c": { + "parameters_hash": "cd658ccd9487f68d0a5c35d3039ac268", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T22:18:53.665196", + "end_time_simulation": 59.97999954223633 + }, + "4d08b6a8d9605e4a936b77b10488e3d8": { + "parameters_hash": "7ee89ad7c7715d692f0f4c3ee0f731b1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T22:18:56.286070", + "end_time_simulation": 59.97999954223633 + }, + "659f33b151294ccb8af928c3c86554b2": { + "parameters_hash": "3cd38d13e901fba838a5ab72ff7ecde9", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:53:35.899485", + "end_time_simulation": 59.97999954223633 + }, + "4d8f57c8487c483cb6cf203c90b6554d": { + "parameters_hash": "2e52dbe254dae0d1234f0a502a882d45", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:37.905868", + "end_time_simulation": 59.97999954223633 + }, + "f86a462866325119ba72f6b229822d01": { + "parameters_hash": "242107598a1a12c70bae99f0680e3548", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:53:40.036634", + "end_time_simulation": 59.97999954223633 + }, + "5592212d0f8e586fa64765f0eba2705c": { + "parameters_hash": "d5e586833f53ee27d99d0daf66dd8732", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:42.172889", + "end_time_simulation": 59.97999954223633 + }, + "e147ef7be54f5563b4c02ab36f1603d0": { + "parameters_hash": "6624c0f700abde074b0ce679986562e4", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:53:44.256952", + "end_time_simulation": 59.97999954223633 + }, + "e8b4c2a4b50959b6b31838fdae59747c": { + "parameters_hash": "05113699d3a2143e8b4ffd2f9ee1c1cb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:46.364689", + "end_time_simulation": 59.97999954223633 + }, + "9192b6981c325c0b91697cb588b4286a": { + "parameters_hash": "5ece1e2f430625602200d5b21b36e60b", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:53:48.755241", + "end_time_simulation": 59.97999954223633 + }, + "3e6deb9875cb50559d8bf854dd5e5dd3": { + "parameters_hash": "e7618b20859970d1f089420557f1b208", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:50.834997", + "end_time_simulation": 59.97999954223633 + }, + "5209ceaff5395d82a77ba8a0ec6d88c0": { + "parameters_hash": "1003314d5d8ed18c3f9d57eee5ca5a9f", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:53:52.856496", + "end_time_simulation": 59.97999954223633 + }, + "f7601fc983584e6fa81bebf4d2687cfb": { + "parameters_hash": "f177901ce797a0b1273bf9de6eb12b03", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:53:55.165240", + "end_time_simulation": 59.97999954223633 + }, + "c9c538896a0146bc98df0c1c4f8e39e3": { + "parameters_hash": "e9f999fcd41603aaa47447fa9c96dd37", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:53:57.408226", + "end_time_simulation": 59.97999954223633 + }, + "5e0289a009de5660a8cb4428d530bd0f": { + "parameters_hash": "10328e606fd117112279a8bea4f3f497", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:53:59.781664", + "end_time_simulation": 59.97999954223633 + }, + "ec65f56742935b379fcea18e6d71efd3": { + "parameters_hash": "c13299df224a1834a8c6c4287cddef25", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:01.865301", + "end_time_simulation": 59.97999954223633 + }, + "8fc49506eed7403f8566c508701eacb2": { + "parameters_hash": "57d8256ee4a5dc2af7bee71ea67efcce", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:54:04.098210", + "end_time_simulation": 59.97999954223633 + }, + "7ca217e06b534e6dbb052614fd02192d": { + "parameters_hash": "2d1db6a8ab621b3a9c9de04883a1e789", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:06.074845", + "end_time_simulation": 59.97999954223633 + }, + "3768ebd518b95856a1fb5e2baef58d60": { + "parameters_hash": "a8eef6e4ffdae94be8e0aa0b507617ed", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:54:08.449716", + "end_time_simulation": 59.97999954223633 + }, + "db44426b4d7c55f09cbf7e0369ed9106": { + "parameters_hash": "ca9550542dd3a22fdf337b541879bef7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:10.462997", + "end_time_simulation": 59.97999954223633 + }, + "b13c8251573c49c9acaf2a5139c17399": { + "parameters_hash": "a5d396bc94cd4ac5234c4918add9228f", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:54:12.685936", + "end_time_simulation": 59.97999954223633 + }, + "f000c5f7e94e4ac4a1edae1353e8c04d": { + "parameters_hash": "0cbc69bdfee4ba057303129b357bfc7e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:14.874066", + "end_time_simulation": 59.97999954223633 + }, + "9e08060fc7bc57d398914826838bfcbd": { + "parameters_hash": "e494be3088ae7626849ed47980de315f", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:54:17.107137", + "end_time_simulation": 59.97999954223633 + }, + "a21a8eccc81a50038058b25c8abe9e1f": { + "parameters_hash": "8015b9b4fbb77d06f6dcaf30e137ec5c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:19.156639", + "end_time_simulation": 59.97999954223633 + }, + "6a22273d32a04da2a7ac789ea4ed92d0": { + "parameters_hash": "154ab511efea5935c632d120b1d6eaa1", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:54:21.421531", + "end_time_simulation": 59.97999954223633 + }, + "da0ebf8b49e0473f9ebca46dd3486766": { + "parameters_hash": "1e93c8ab8c1e7b223b384414a9d5e23c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:23.477649", + "end_time_simulation": 59.97999954223633 + }, + "0f59b94bb99c5bbf984cfc9e2f77ed73": { + "parameters_hash": "86dd972325c92aa23efd4a1ac614f9d0", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T22:18:59.077172", + "end_time_simulation": 59.97999954223633 + }, + "75f78e0515645f3d8659c29ef5e14aff": { + "parameters_hash": "1c01cb8ffdbb20a6b637552a75ba8b74", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T22:19:01.606402", + "end_time_simulation": 59.97999954223633 + }, + "8d48963773a554a7a701cf6bf9cc0461": { + "parameters_hash": "b37e0fe59539de1f9d2ad41c2516d8c0", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:54:25.758174", + "end_time_simulation": 59.97999954223633 + }, + "9683b8726e0f520899b7eb072eeec879": { + "parameters_hash": "979210640384b0bbbdf4cd1c25447dcf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:28.107293", + "end_time_simulation": 59.97999954223633 + }, + "cc6320c44b045225b9eae9b66af2a2bb": { + "parameters_hash": "ecfc0184533c36ad352500910eeb9eb4", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:54:30.384850", + "end_time_simulation": 59.97999954223633 + }, + "506904b20a635080b10b812a65b9c6a7": { + "parameters_hash": "0438aaecca5594f530f549337e670ab4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:54:32.469269", + "end_time_simulation": 59.97999954223633 + }, + "a9b3ea901e82539ebd2c6e4ad10e0f1a": { + "parameters_hash": "d1488815ce482f31cee7bc43db655f38", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T15:10:01.264474", + "end_time_simulation": 59.97999954223633 + }, + "99ea1c39535b4869bdf816c53fef8829": { + "parameters_hash": "c2ff93c8fe59be6517e1b7c437506285", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T15:10:04.233468", + "end_time_simulation": 59.97999954223633 + }, + "4a33f63ab9454f5f80585ba258ecd353": { + "parameters_hash": "9cf0668670d92000e7a99ededed647b6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:06.921678", + "end_time_simulation": 59.97999954223633 + }, + "a95beb9c295d5f368324edb63f13ed44": { + "parameters_hash": "d973e7b8ad2f15f6f0f8773762c28a4a", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T15:10:09.785904", + "end_time_simulation": 59.97999954223633 + }, + "8944adad6b5559af9f462be2358ce3fd": { + "parameters_hash": "d9fe8bcbd8725cd703f08b0fb92b3418", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:12.473058", + "end_time_simulation": 59.97999954223633 + }, + "a3ebd4caf5665f29a8dabef5b17fff49": { + "parameters_hash": "1c4b80b06595d9a8de90f209ce299a82", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T15:10:15.483072", + "end_time_simulation": 59.97999954223633 + }, + "2e4d8a7328a6506ead22a8e0ce487782": { + "parameters_hash": "39195dd5250927b21bd54006dfe2c52f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:18.072296", + "end_time_simulation": 59.97999954223633 + }, + "c830b79b9b444a229f5ed1f529f70904": { + "parameters_hash": "96fdb6e6444a98f343908a7d2fb1acbe", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T15:10:20.508576", + "end_time_simulation": 59.97999954223633 + }, + "54ebfbff91274eb9b5a3dc5a7be595f2": { + "parameters_hash": "2912f0e88c550f638d880ec6bf033a59", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:23.050554", + "end_time_simulation": 59.97999954223633 + }, + "59cbf6623d1848b1a7fa0a6c11886363": { + "parameters_hash": "24f036170a4e8667feb37bdebcfd8795", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T15:10:25.814651", + "end_time_simulation": 59.97999954223633 + }, + "cf5b2f11716643cebd91e39f0dbbdad4": { + "parameters_hash": "5e59e3239e1922b844629aa9edcab014", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:28.505064", + "end_time_simulation": 59.97999954223633 + }, + "041f71009f01531e9eb36542b97897e8": { + "parameters_hash": "d42e930a2bc0cb74663b80f5756d0429", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T15:10:32.062315", + "end_time_simulation": 59.97999954223633 + }, + "62d788681231544194c2d089cc60ae5d": { + "parameters_hash": "ce374b08b75946d3b0e7dcaa7b38257e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:35.168610", + "end_time_simulation": 59.97999954223633 + }, + "df17049fa32b57e790dca8f8139406f7": { + "parameters_hash": "a943c9c5e7851e345accf2e381ed653e", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-04T15:10:38.441693", + "end_time_simulation": 59.97999954223633 + }, + "5001e738e3a351da81d59e18937bb404": { + "parameters_hash": "b6d2539996817170ed9f1253fdf12d4c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:41.543184", + "end_time_simulation": 59.97999954223633 + }, + "9881fa6ffe12536c8817c68aa68194cc": { + "parameters_hash": "fb5543829a0de9ee346730675c2e0ebf", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T15:10:45.668676", + "end_time_simulation": 59.97999954223633 + }, + "1982a1aeee245c1bba379111099c39c1": { + "parameters_hash": "345e3e918340a256c26cda34161fa9dc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:48.474365", + "end_time_simulation": 59.97999954223633 + }, + "46300f05f82a518796cc12c2737683f8": { + "parameters_hash": "5042286529695b6464964a20034f2335", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T15:10:52.777288", + "end_time_simulation": 59.97999954223633 + }, + "020bfb4d384c519c812e93b10bd63506": { + "parameters_hash": "2e067566131a1cd78359a638d1b8679b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:10:56.179158", + "end_time_simulation": 59.97999954223633 + }, + "751bf183c73c540cb453f8565567b2b9": { + "parameters_hash": "7ccee84bc077cafef8d15731e4c2909e", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:55:14.788544", + "end_time_simulation": 59.97999954223633 + }, + "361faf02139046e893bc427b21fc52ff": { + "parameters_hash": "14d272d28a40d0a9b85e0c9b180ef8e1", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:55:17.472492", + "end_time_simulation": 59.97999954223633 + }, + "066f20c236094f769c038ade0c4ac306": { + "parameters_hash": "553a28e11bf04d4851690f992e7dc767", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:19.591380", + "end_time_simulation": 59.97999954223633 + }, + "a9e3bfc1a91e5fe3b265677f665e0282": { + "parameters_hash": "726b6476bf8d1b52874c02e6a155839a", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:55:21.999167", + "end_time_simulation": 59.97999954223633 + }, + "3f1bbfae5174557ba6fda2283a896227": { + "parameters_hash": "819c72198dfff6b47e40412ba04984e5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:24.664702", + "end_time_simulation": 59.97999954223633 + }, + "bebb77d0250e5599966348a2c854c5ce": { + "parameters_hash": "9b49a9306442068c2620a8f61044fd33", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:55:27.214437", + "end_time_simulation": 59.97999954223633 + }, + "c73786fbec6d5308874f327ba2d79078": { + "parameters_hash": "aa7e8a5aac2f2a6dcc8ff31b7fa99068", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:29.219935", + "end_time_simulation": 59.97999954223633 + }, + "7b4e20ff78014b1bbcada71d5bf4c1b9": { + "parameters_hash": "c9f8b165558ebadf7ab368e7403acafd", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:55:31.550607", + "end_time_simulation": 59.97999954223633 + }, + "2169125f3567482a89823885b9add497": { + "parameters_hash": "eb245feef2fa6af0bbb90e3bf19267f1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:33.642001", + "end_time_simulation": 59.97999954223633 + }, + "f883d4d7ec895060acd55a0779d15888": { + "parameters_hash": "353709bbdb4c4c34949a6dcc3989929f", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:55:35.849705", + "end_time_simulation": 59.97999954223633 + }, + "97cedd0220dd5557803c52afce3a3895": { + "parameters_hash": "bd6e82d1eb004ed2a9ed2a692af733fc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:37.990611", + "end_time_simulation": 59.97999954223633 + }, + "5774aa02a7c4456fb660a99d016fcb7d": { + "parameters_hash": "d44031c415e188eca1038c1febae1eca", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:55:40.135538", + "end_time_simulation": 59.97999954223633 + }, + "f5a995dd858545348384770bfa89521c": { + "parameters_hash": "b66fba67646ede50f65cdd654d3d0f7d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:44.702996", + "end_time_simulation": 59.97999954223633 + }, + "f11188670cd55f239dc2d21c1a12ca9a": { + "parameters_hash": "3656e5486e967e2be7616733b2bb0da2", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:55:46.824207", + "end_time_simulation": 59.97999954223633 + }, + "ead7dc2f2b145a93a617fa4abc250d9b": { + "parameters_hash": "fe2e925a907a3afa78b3a69923934e46", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:48.928868", + "end_time_simulation": 59.97999954223633 + }, + "ce53b3549a4857acbcc24fd0583720a2": { + "parameters_hash": "d87f8fc53a175c50ee3d9d777e3fdfad", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:55:51.221412", + "end_time_simulation": 59.97999954223633 + }, + "5885b088ef9e5512a16b40c69d01f3b0": { + "parameters_hash": "ace9b3b3db0b89b3558d6df903b197c5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:53.276511", + "end_time_simulation": 59.97999954223633 + }, + "945c5a0b7352554f87e4ee70d573a63e": { + "parameters_hash": "7790bc81eef5f820937232a966f6ad08", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:55:55.556273", + "end_time_simulation": 59.97999954223633 + }, + "14acaa1d82d25332a209397af30e50c7": { + "parameters_hash": "f5605c02fa74c511bf79c242802f99ac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:55:57.522253", + "end_time_simulation": 59.97999954223633 + }, + "7ceb057bc47e58eeb49d59b1f706c140": { + "parameters_hash": "8d02583883122be2010ee5fc6fe65ea7", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T15:10:59.379290", + "end_time_simulation": 59.97999954223633 + }, + "ffab94d145a54c32a63de067e2d95f7f": { + "parameters_hash": "5a78f7cd1d97bb0c6b417148498dabcf", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T15:11:02.363772", + "end_time_simulation": 59.97999954223633 + }, + "7308dcb2eb224ee5aab6ed4aadce5292": { + "parameters_hash": "9f4ed503bbc67c76e65216b3f266b01e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:05.461061", + "end_time_simulation": 59.97999954223633 + }, + "5e7ee865a19a5fe89c8f48739ac24c45": { + "parameters_hash": "c0767067d8daf8df907244e92fe75e33", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T15:11:08.516965", + "end_time_simulation": 59.97999954223633 + }, + "415b1d8e847e5e42845db38bda8b4a83": { + "parameters_hash": "6dc6dc191a63fbac9e9bb59dc25e41a9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:11.136827", + "end_time_simulation": 59.97999954223633 + }, + "132fe47a7ebe5a0590dcfce5b44b9be4": { + "parameters_hash": "099f8aaeb1d0a975a63fc0d37e3d52e5", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T15:11:14.360819", + "end_time_simulation": 59.97999954223633 + }, + "76e9266788e458e98803b76256bcf69f": { + "parameters_hash": "480c78ea7b8b3664b89783e9521c207b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:17.259796", + "end_time_simulation": 59.97999954223633 + }, + "2286dbde602d43658e2cbcf336930b72": { + "parameters_hash": "d4bf7321cf0ca79f9be144309cf45305", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T15:11:20.175992", + "end_time_simulation": 59.97999954223633 + }, + "7fc9b77c2c9d42dbbec39a28e5147eb8": { + "parameters_hash": "301a73b6aa02acc7dc350dbae287aa01", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:22.688756", + "end_time_simulation": 59.97999954223633 + }, + "651da37585894e82988296f28bd55c1d": { + "parameters_hash": "71e9b95c3a283fdefcad4dd136bb0c8a", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T15:11:25.603593", + "end_time_simulation": 59.97999954223633 + }, + "15cb8274b5f348b0aded8d10a6fe0cf7": { + "parameters_hash": "071fad3dca0544201fc1182deae6aff9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:28.102738", + "end_time_simulation": 59.97999954223633 + }, + "0662acd660d859078cce3dd368e1d793": { + "parameters_hash": "c78c8c03366356d712f818177d911441", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T15:11:31.110671", + "end_time_simulation": 59.97999954223633 + }, + "8c1bdb40c82650eabe1410971757e06f": { + "parameters_hash": "4ffd3a126c59ff45043c57ecd1b13510", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:34.426860", + "end_time_simulation": 59.97999954223633 + }, + "c310ff8b768541e6b7c059b574c697f8": { + "parameters_hash": "3f6cfd6e5e3299aa7a8970c0d9de7f34", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-04T15:11:37.313322", + "end_time_simulation": 59.97999954223633 + }, + "7ce4959ab74545f2a99fac1f9c9e190b": { + "parameters_hash": "480416eea16ec3ff37a8f3d622662e8d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:40.439156", + "end_time_simulation": 59.97999954223633 + }, + "e3bf6afe41a75f7ca9faca331602edf3": { + "parameters_hash": "2008cc2a4290241644c5bcd0f9c90efa", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-04T15:11:43.531912", + "end_time_simulation": 59.97999954223633 + }, + "401d629458e85093af379e6c403c7bd8": { + "parameters_hash": "4eefc3467c83b18b9f16f610d7e6c68e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:46.708161", + "end_time_simulation": 59.97999954223633 + }, + "8da30893b98f57c7844ae2e2941687da": { + "parameters_hash": "d0ce67fbc7c42c722909f81f47db2d6f", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T15:11:49.580535", + "end_time_simulation": 59.97999954223633 + }, + "5d65e1dfa5295f4784e1dbc2b0c1e23d": { + "parameters_hash": "08816d095c9b1378ddfde52307350964", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:11:53.104317", + "end_time_simulation": 59.97999954223633 + }, + "f128e30cbf385513bd576ed7456e2a6d": { + "parameters_hash": "9ea90b6cd842b17545676d410539b049", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T15:11:56.873677", + "end_time_simulation": 59.97999954223633 + }, + "22a153c5c92150518e20423beaa9dd42": { + "parameters_hash": "8b5139168a49bc20258e3bb36b3a9f69", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:12:00.286044", + "end_time_simulation": 59.97999954223633 + }, + "da13aa0075ca590bb250eb0403648b84": { + "parameters_hash": "ef89f08e0482a4b619c39445ac3d5ce8", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T12:56:32.495178", + "end_time_simulation": 59.97999954223633 + }, + "561af90c1b354f54acbad9f0704000c0": { + "parameters_hash": "ffb323242ed947da7216f5eafac55017", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:56:34.707273", + "end_time_simulation": 59.97999954223633 + }, + "07b66a32286a42b4829ec86e091bb545": { + "parameters_hash": "ff0c1eb51e2bae3c9e5dc197768e661c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:56:36.935138", + "end_time_simulation": 59.97999954223633 + }, + "1b4be053be5959d2ae8211c077a32bcf": { + "parameters_hash": "b2fc8739e35c853efe36e3b096e827fc", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T12:56:39.081237", + "end_time_simulation": 59.97999954223633 + }, + "2a9ec50092c351718ebfb16f6ca05089": { + "parameters_hash": "f45c2f90bea7f96ced477eb98119aa90", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:56:41.190274", + "end_time_simulation": 59.97999954223633 + }, + "0d4cf60257774d9ea23cf538a161d723": { + "parameters_hash": "9cc54e1564bfdac7be35206535be375f", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T12:56:43.355677", + "end_time_simulation": 59.97999954223633 + }, + "b502923105764d42825815a9e8a5462c": { + "parameters_hash": "8395a4503ea1936754fb237788bcb78b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:56:45.422966", + "end_time_simulation": 59.97999954223633 + }, + "201645029428467fa5a5f5722f39ff9f": { + "parameters_hash": "fc9b39dd784825ef81de3e46ca98547a", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:56:47.618776", + "end_time_simulation": 59.97999954223633 + }, + "27b815f8181840419195f0cb9102f1c9": { + "parameters_hash": "4bbe95ddf158644d2732cc0667bf9913", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:56:49.775051", + "end_time_simulation": 59.97999954223633 + }, + "1279d63b27cd587eb34d44748bf4a83e": { + "parameters_hash": "85a3b5bf6affc3d7871f387df98d3a3d", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T12:56:51.924220", + "end_time_simulation": 59.97999954223633 + }, + "d1bad042622c501d804c05e68a64ed53": { + "parameters_hash": "04a1353c6b719dcf429fd9b0a67f3142", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:56:53.947306", + "end_time_simulation": 59.97999954223633 + }, + "ed47390d6b32503bbc393b0595e2cf79": { + "parameters_hash": "2b99afad18af0ea3ee92804b95ca5799", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T12:56:56.294630", + "end_time_simulation": 59.97999954223633 + }, + "90a412f5679c5872a2b55309d8988b1d": { + "parameters_hash": "5b13c73af441f12efa23624458eaa536", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:56:58.470393", + "end_time_simulation": 59.97999954223633 + }, + "245e2f74390953df812061e150fcd476": { + "parameters_hash": "94e294767953f3dbb06d63983effcc32", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T12:57:00.591103", + "end_time_simulation": 59.97999954223633 + }, + "2f39f50a4b5d53a0b4493894c67e5cb4": { + "parameters_hash": "2221acbdd0446b863504edd330e5778c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T12:57:02.671655", + "end_time_simulation": 59.97999954223633 + }, + "87668905ce9a59c69900b28efc7e97a1": { + "parameters_hash": "b899d83efe23edaba105891c38592c8e", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T23:10:38.029208", + "end_time_simulation": 59.97999954223633 + }, + "eef23a2370785b99aa4438937609d4e2": { + "parameters_hash": "8feb26b38b8504c3c36288dbe3cd8cfe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:10:40.616833", + "end_time_simulation": 59.97999954223633 + }, + "56b2f1a7b7895af6ad769243c1944e79": { + "parameters_hash": "00929b9d92ac811dd747b4742f73e0b6", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T13:52:39.242823", + "end_time_simulation": 59.97999954223633 + }, + "91bd13d01bb54bc08e1aad8c7208c42d": { + "parameters_hash": "385cee851beb986fe105fe43af219345", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T13:52:43.012074", + "end_time_simulation": 59.97999954223633 + }, + "40c25fa718be4d1a85cf4db66a419826": { + "parameters_hash": "85308b8bed9f7af267f46cdbe0068afa", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:52:45.816598", + "end_time_simulation": 59.97999954223633 + }, + "bedf71dbd54d5b159500af85a481f2ca": { + "parameters_hash": "c484a4f5c79f7021734a994b42ee31cf", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T13:52:49.231021", + "end_time_simulation": 59.97999954223633 + }, + "e15ca30b15e1587db16fb3a8a483db3c": { + "parameters_hash": "8c432ea37c8d05cf74baddc7e197f332", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:52:51.765512", + "end_time_simulation": 59.97999954223633 + }, + "879981a7651953eab7d7b77c638befbc": { + "parameters_hash": "1c91f26464d5a0569f612a9f058f3ddf", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T13:52:54.979558", + "end_time_simulation": 59.97999954223633 + }, + "32ff4c3f892d5368b6f5ee2006d2c666": { + "parameters_hash": "56fb44a5050c1b358e2b30727cbe85d1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:52:57.904095", + "end_time_simulation": 59.97999954223633 + }, + "f51752f943a641a4980e0bcb81b13a25": { + "parameters_hash": "0ffeb3da9dc4c96e85cdfdd27b9e92b8", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-04T13:53:00.947536", + "end_time_simulation": 59.97999954223633 + }, + "da281fb32cb24eca92d2616d70dddddf": { + "parameters_hash": "2e50d1dcd7caf895e1a9e616b4cf9c30", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:53:04.496213", + "end_time_simulation": 59.97999954223633 + }, + "980d1ab787984d38beb8bec7989c0fab": { + "parameters_hash": "3c81ef1ceafad8075a1f11e0565150b4", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T13:53:07.956304", + "end_time_simulation": 59.97999954223633 + }, + "ea7b761558de4a878315d14167895947": { + "parameters_hash": "aeda92b3d2d5994da9bcfd34ac251506", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:53:10.707176", + "end_time_simulation": 59.97999954223633 + }, + "e4892dd45cad5af4bcf4351e8937807b": { + "parameters_hash": "b9931f99de97ac005a125b4c97e82b27", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-04T13:53:14.023109", + "end_time_simulation": 59.97999954223633 + }, + "53374605aed55526a8a9133389611951": { + "parameters_hash": "73c4d4c1f9f406c346fe0e690cadb44c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:53:16.461994", + "end_time_simulation": 59.97999954223633 + }, + "d78bfb68a41f406687c434b3742bcafa": { + "parameters_hash": "18d7920c3d3d363bbf9752d45be6adab", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-04T13:53:19.785793", + "end_time_simulation": 59.97999954223633 + }, + "bbbe8ac29c4440a99615eb4398c82c86": { + "parameters_hash": "5f452d973efa3524a1a2951268bb2cc1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:53:23.357742", + "end_time_simulation": 59.97999954223633 + }, + "86e5724b2388549fa5c2a2dc4440a226": { + "parameters_hash": "7381c17257d6fa57d611876239cafab5", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-04T13:53:26.904681", + "end_time_simulation": 59.97999954223633 + }, + "69f59f52fa6058ec9ff3c57c5be06885": { + "parameters_hash": "e50b4762dddf8551ccec9c0723b08151", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:53:30.037275", + "end_time_simulation": 59.97999954223633 + }, + "dce82ef055ad5132adadeb3d2b83a6ae": { + "parameters_hash": "f22d4b213d64167801fcf811704c31ab", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T13:53:33.380594", + "end_time_simulation": 59.97999954223633 + }, + "40037c9332055257b95af5c4395cea7b": { + "parameters_hash": "b978d350057d743ccf69f2fa5545d371", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:53:36.463900", + "end_time_simulation": 59.97999954223633 + }, + "92e6cd064ff6505e89f37f4b55e77043": { + "parameters_hash": "8ab9af1415e6c9ae006f7f6da52714b8", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-04T13:53:39.692458", + "end_time_simulation": 59.97999954223633 + }, + "17b84a061a2754848390c3f674dcaf45": { + "parameters_hash": "d894da68dccb64769e81151ff8a4d306", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T13:53:42.840255", + "end_time_simulation": 59.97999954223633 + }, + "20bcea5c4c6259eb913c324afcffc7ab": { + "parameters_hash": "f5e2c3c06a0d59d972ef1c06acaf85b8", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-01T23:27:03.252595", + "end_time_simulation": 59.97999954223633 + }, + "7b91f6c5e3c342d794c5baf4a740e3b7": { + "parameters_hash": "d8f1e07cdfaa1fc5f48c7357f3501942", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T23:27:06.315539", + "end_time_simulation": 59.97999954223633 + }, + "3064b98a61e6401c8256ff50cfe83ab1": { + "parameters_hash": "b2e6df1f61c0ba1d5b4cd25b75528a0d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:08.999359", + "end_time_simulation": 59.97999954223633 + }, + "616a9586fe765e6f9c92db29946198bb": { + "parameters_hash": "0240c3e4773f763bfd28c70fd845e5de", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-01T23:27:12.192377", + "end_time_simulation": 59.97999954223633 + }, + "5586366bf3865b269fe14845fa6a8892": { + "parameters_hash": "c8a69356db43a15e2f3d51d09c6dbbb2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:14.749134", + "end_time_simulation": 59.97999954223633 + }, + "7fda0f84bf684b3e8f1f4acc2b71a87c": { + "parameters_hash": "6527ad45f4784e66bc960329e8bc2903", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-01T23:27:17.401230", + "end_time_simulation": 59.97999954223633 + }, + "99a28b7df6be4655a82d9d7b45be39d4": { + "parameters_hash": "039cf659c0e25f0b8749fb4d1c2426d7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:20.332295", + "end_time_simulation": 59.97999954223633 + }, + "718c1f8d7798591db3f76c8dac03fb96": { + "parameters_hash": "2844700e755081976d2c86a45f540968", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-01T23:27:23.318867", + "end_time_simulation": 59.97999954223633 + }, + "822bf75e08af5c30952fb6b2f09d02c2": { + "parameters_hash": "35f28cbb29a02c3148dd98449483ae1f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:26.448873", + "end_time_simulation": 59.97999954223633 + }, + "ae1ca2945b6443abaabd3974f08fc4c3": { + "parameters_hash": "8e0c511c35d3a277af03ed0c85be088c", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T23:27:29.324836", + "end_time_simulation": 59.97999954223633 + }, + "b7690ca490ad486183cb3e82758b85ac": { + "parameters_hash": "ea4721579465f7793b46299292713828", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:32.271767", + "end_time_simulation": 59.97999954223633 + }, + "c1c17238711356fc8d868228147331a8": { + "parameters_hash": "178f27175bdd92468f1884acb57f8a08", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-01T23:27:35.198031", + "end_time_simulation": 59.97999954223633 + }, + "ef6cff31a49a5e0e821f1ddc9601829c": { + "parameters_hash": "22bb5d076863e08648d82b2521025ce9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:37.687908", + "end_time_simulation": 59.97999954223633 + }, + "eaad545c18fb52ca838cc86b326c432e": { + "parameters_hash": "0067c5df4b631fa3fbc1534405233bfc", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T23:27:40.843728", + "end_time_simulation": 59.97999954223633 + }, + "70e7e8f9d12d515392b1808984b036e2": { + "parameters_hash": "5bea069da9013b56692a923e1af0b8cd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:43.899635", + "end_time_simulation": 59.97999954223633 + }, + "23d5a3a8ef175d13ba8084fa93b0c387": { + "parameters_hash": "d1b79c3f13298f16d61cce74d9a17c6c", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-01T23:27:46.964225", + "end_time_simulation": 59.97999954223633 + }, + "9c5390b312bd5ec980d51ac783abdc99": { + "parameters_hash": "f57def3a30a00c2446faa2345540e6bd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-01T23:27:49.568337", + "end_time_simulation": 59.97999954223633 + }, + "cc800d49ccb8519788e0fba73cf951cc": { + "parameters_hash": "99453b87b10d308afc123c5bf0d08935", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T19:59:01.404832", + "end_time_simulation": 59.97999954223633 + }, + "0e90da8030f55dc9a513525ae819286a": { + "parameters_hash": "4aa9968b070128c293db9b7c57f703d9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:59:07.253801", + "end_time_simulation": 59.97999954223633 + }, + "041d6bafcb8f55729d1f126d6a507afb": { + "parameters_hash": "d3d73b63d624c880b8cc9e415e8f0d2d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:11.192703", + "end_time_simulation": 59.97999954223633 + }, + "c4d8b5eeb6a75ec6a9aa21df3d821cef": { + "parameters_hash": "b759d3c7f1d9ead217c2dcc1ab56aa85", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T19:59:14.215030", + "end_time_simulation": 59.97999954223633 + }, + "bc73ce4ba6bd511aade3d16b4d245e6e": { + "parameters_hash": "0d70806da5ef26e042a8efc0ef47fbf6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:16.985225", + "end_time_simulation": 59.97999954223633 + }, + "6e739c92001552d1bf0537222af401da": { + "parameters_hash": "ebaec3de8201827f650b95eff2905f37", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T19:59:20.588915", + "end_time_simulation": 59.97999954223633 + }, + "766bbecdbdf95e7482554287dcef5649": { + "parameters_hash": "535dbd5255817f51ccd637fd3eb3d6dc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:23.441812", + "end_time_simulation": 59.97999954223633 + }, + "7df93de035805e598e695b95e94172fd": { + "parameters_hash": "4451055a3114a7e5f66ba23039ed918f", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T19:59:26.228108", + "end_time_simulation": 59.97999954223633 + }, + "f53fb7543a065f1cb5c808b71c44c6e2": { + "parameters_hash": "2beec5335ecc2265b3b1c2a2a51834f5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:29.463266", + "end_time_simulation": 59.97999954223633 + }, + "6dd72fab1bff539eb989c32f645fe917": { + "parameters_hash": "6a2f83661c020a596195cbb7e29c0eec", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T19:59:32.446994", + "end_time_simulation": 59.97999954223633 + }, + "68912d9930be5a3898b980be1dcd2a41": { + "parameters_hash": "5c38dac1eb67fbd84639b15b16afa63d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:35.265613", + "end_time_simulation": 59.97999954223633 + }, + "61e9b2289fd05fb2a3b0a90456156ade": { + "parameters_hash": "03229ac833427ac7ef9730b0728b7130", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:59:38.681302", + "end_time_simulation": 59.97999954223633 + }, + "c8f184b77560543bb7c9170fd35c523f": { + "parameters_hash": "220d37746a000f65b5145e0cfd59f93c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:42.060757", + "end_time_simulation": 59.97999954223633 + }, + "ec6ca53029d75d7089109d75fdaffccf": { + "parameters_hash": "bd780e9f1d18e3d3ba4a28e72dbc4774", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T19:59:45.064851", + "end_time_simulation": 59.97999954223633 + }, + "5729efbefd37515194aa4d932ace7bd9": { + "parameters_hash": "5184b77995a87d15720f634371866847", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:48.084488", + "end_time_simulation": 59.97999954223633 + }, + "1ec3e6bc4d275d74917d24cda8f7d69b": { + "parameters_hash": "f94addefe3c86c0c1088930521885ac7", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:59:51.075067", + "end_time_simulation": 59.97999954223633 + }, + "f1e2a9f68ce352da8e6d37c177f0d031": { + "parameters_hash": "4fcc1c42aa7be0f4b452a1f77a33be2f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:59:54.152517", + "end_time_simulation": 59.97999954223633 + }, + "09ce4262dccd56af82f3a9c958862a00": { + "parameters_hash": "b88590f3429eba5ee21d2d79b738f389", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T19:59:57.440164", + "end_time_simulation": 59.97999954223633 + }, + "8f59a7fe12f85e37ad029f5fd9944f23": { + "parameters_hash": "ff374286ba477d0492e40cbe88b1764d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:00.312252", + "end_time_simulation": 59.97999954223633 + }, + "adc34ff7af335a71aa6ac5609136b4c6": { + "parameters_hash": "a76040bb3b9887b3a65b2db4d1f6109e", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:00:03.038878", + "end_time_simulation": 59.97999954223633 + }, + "cf75e495461d55d8aa5ec28041d1eba4": { + "parameters_hash": "3664bee5b63f55972e633fdab0f2f4de", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:00:05.923846", + "end_time_simulation": 59.97999954223633 + }, + "2c10b8c4b4ae58f780302d9eadaae54a": { + "parameters_hash": "fed036e38b20bdd7a80b34b1435fc30a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:08.537309", + "end_time_simulation": 59.97999954223633 + }, + "1f5e108b1de457878df3dfcf6063bc90": { + "parameters_hash": "ecb4922b7cd6948c8f1cba372fe13c90", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:00:11.761361", + "end_time_simulation": 59.97999954223633 + }, + "28d390b1b8ce52c0a1806f495a9016ce": { + "parameters_hash": "2148d5a324adf4b9449d627a17d0bb72", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:14.794236", + "end_time_simulation": 59.97999954223633 + }, + "9d663cc7c6ba51b9ac6ea0adca75b204": { + "parameters_hash": "2cca6c19707debf98d7873d5ba243bfa", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:00:18.081679", + "end_time_simulation": 59.97999954223633 + }, + "ca224a1326b957eb94e90603bf4fc209": { + "parameters_hash": "97cab78b3a95a67d1730bc06fdf03299", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:21.006992", + "end_time_simulation": 59.97999954223633 + }, + "77ba0ef646195a739efc5cfd1ecf224f": { + "parameters_hash": "d07fd271bb34bc4c9c002bdbf9121d35", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:00:24.127436", + "end_time_simulation": 59.97999954223633 + }, + "e88586f6f51a5bd38188298079ded905": { + "parameters_hash": "ddac9538ba0d79d4fffeb47f742643c7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:26.719419", + "end_time_simulation": 59.97999954223633 + }, + "d066da1c02e35ef3b84b231211d6a317": { + "parameters_hash": "ac2b6407f3b9517582785b35b52ad701", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:00:29.571321", + "end_time_simulation": 59.97999954223633 + }, + "4c0fb9c925ba58b3bdc7b026674c179b": { + "parameters_hash": "9a2d48279d838ae300bc55703e356268", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:32.217111", + "end_time_simulation": 59.97999954223633 + }, + "f77d008bc32e5531815470dae51cf67f": { + "parameters_hash": "e9f059561d19edf9e934c12dc2364a03", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:00:35.248847", + "end_time_simulation": 59.97999954223633 + }, + "5fe5f9981eea52f08d788303e7de8bb9": { + "parameters_hash": "873770835e39bec2229dc060ef0d8e94", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:37.783983", + "end_time_simulation": 59.97999954223633 + }, + "43dfeb598cc654f8903c98baebfbed31": { + "parameters_hash": "e43ad82c19427cca8c14828e15d19f15", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:00:40.669657", + "end_time_simulation": 59.97999954223633 + }, + "91d2a0e3b23a542cb6bb2de3c801ea97": { + "parameters_hash": "a2018c066d352a07a26a7aab48b2c01b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:43.033704", + "end_time_simulation": 59.97999954223633 + }, + "003f425c768e5f189f40e1b0b2f3ca3d": { + "parameters_hash": "316973eddf32f426a3b2760424f843e7", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:00:45.760869", + "end_time_simulation": 59.97999954223633 + }, + "5e3b1b0c765156f7b01b339fa1202f64": { + "parameters_hash": "71234cd6d0b75007ae2fbe0399ddc611", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:48.690093", + "end_time_simulation": 59.97999954223633 + }, + "fab54408482f5222950d829ecec9502c": { + "parameters_hash": "0d9d02f4d9e910e420989c776a344a61", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:00:51.716372", + "end_time_simulation": 59.97999954223633 + }, + "f27349e100d854c59ebba8df2164aefa": { + "parameters_hash": "89df880b288e25250b7a134c3e7d4c65", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:00:54.143414", + "end_time_simulation": 59.97999954223633 + }, + "bb81df685f9a58e59ecb171a9cceb33e": { + "parameters_hash": "402af21e1c849e04ac412fe2d544fbf5", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:00:56.891061", + "end_time_simulation": 59.97999954223633 + }, + "b8b64d9c72a35d9486886e9b726261cf": { + "parameters_hash": "fd447dc95cdda67db7991144d161c076", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:01:00.189331", + "end_time_simulation": 59.97999954223633 + }, + "5585abad4ea65a158b77b3071e22c0e5": { + "parameters_hash": "b1a085d6d4e38c0de14e7d45999a0638", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:02.784797", + "end_time_simulation": 59.97999954223633 + }, + "343f4eb4cc3c50beafd187dcfe319f27": { + "parameters_hash": "e28b590bd1660d635ad0f479e3e7a689", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:01:05.583988", + "end_time_simulation": 59.97999954223633 + }, + "59fafb688b0b5928a7179bdcbd472e5d": { + "parameters_hash": "ccdc61cf4a0b8c62a50ca5ae43d3a9e9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:08.073761", + "end_time_simulation": 59.97999954223633 + }, + "5eae2d4c6c535d67bc529d983994c536": { + "parameters_hash": "5e0f8c6751aaf6de6547de65a91544a8", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:01:11.044486", + "end_time_simulation": 59.97999954223633 + }, + "7b1d763720975e97b2364246746ef69d": { + "parameters_hash": "193079c0f40194281cb375f30f598152", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:13.809044", + "end_time_simulation": 59.97999954223633 + }, + "c48a3b8ec2d55d168f71118045fee508": { + "parameters_hash": "92bb950cc57452fc961e46ec59339ff2", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:01:16.316491", + "end_time_simulation": 59.97999954223633 + }, + "54566b621e1f594eb1ba6c11ad8fc7c2": { + "parameters_hash": "ec5d134828471a8350b0c7f1e7beefe5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:18.726091", + "end_time_simulation": 59.97999954223633 + }, + "c32fcada64fe5036a1a843f7f423f6b4": { + "parameters_hash": "8e4b8dd31d946bf8483cb98cc6ca28f4", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:01:21.366501", + "end_time_simulation": 59.97999954223633 + }, + "f9b3f1ed029e557ea5a7bc669907baa3": { + "parameters_hash": "9e31495032951dffeb09d6dd52572c69", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:23.774348", + "end_time_simulation": 59.97999954223633 + }, + "ca738e11203c529eb8d7ac37d0ce06da": { + "parameters_hash": "e65f1f2b7ed7c9fbd3959964b3380d20", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:01:26.392392", + "end_time_simulation": 59.97999954223633 + }, + "7cbd112d1bee58b69ccfdec6869ac764": { + "parameters_hash": "ac314a1c66e8bd6f48ae7bf82b1a378b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:28.860808", + "end_time_simulation": 59.97999954223633 + }, + "f338946e9bae571b8c168fea572107c9": { + "parameters_hash": "9be726c39653c602f695086eccda531f", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:01:31.588047", + "end_time_simulation": 59.97999954223633 + }, + "dcdda4eee3795725a2f5feeaffda6580": { + "parameters_hash": "6f1e421e5c3cac72ef6e4340f0a06c96", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:36.364832", + "end_time_simulation": 59.97999954223633 + }, + "9a03fccac5e1443fbea77e7243b73b84": { + "parameters_hash": "63687f5fc9371aefd313ed10992bd199", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:47:53.948955", + "end_time_simulation": 59.97999954223633 + }, + "bf1eac940d605849a6fb3a7d048864da": { + "parameters_hash": "ece2b41e16ca4ea372fccd05cf6183e3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:47:57.284559", + "end_time_simulation": 59.97999954223633 + }, + "a0a52b8d35e750c8ad5b9b03de916742": { + "parameters_hash": "db1a8fff74182913f0e936a374e0af05", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:01:39.438079", + "end_time_simulation": 59.97999954223633 + }, + "1a70b12d7dcc5e28b4596670f09e61a6": { + "parameters_hash": "9b1c09d99b330a405169805872432e02", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:42.027786", + "end_time_simulation": 59.97999954223633 + }, + "f3c6b991aba15df2ac2d1c314d4249f5": { + "parameters_hash": "cbc35e1abf06e75c543671dc9bd29b7c", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:01:45.056910", + "end_time_simulation": 59.97999954223633 + }, + "bb645aa489b75946b343e3bc3eb93fff": { + "parameters_hash": "c2ffcf83cb39de138f0887f6557abf85", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:47.720312", + "end_time_simulation": 59.97999954223633 + }, + "f1076dc3e4ff45b685096d28d47d4b70": { + "parameters_hash": "fc44b0b90286791adf6187c6c53c39d4", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:54.250628", + "end_time_simulation": 59.97999954223633 + }, + "c936db1419db5f168db4e0a50ad54242": { + "parameters_hash": "e502a6cc55e4264958b31e140fa12d8b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:52:56.759143", + "end_time_simulation": 59.97999954223633 + }, + "25929fbc910e457d9143e2899419263b": { + "parameters_hash": "da810cf9ce67bbdc66427b5a54425d8d", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:52:59.648568", + "end_time_simulation": 59.97999954223633 + }, + "8d5954ad700a5b6ebb6dac20de0a0efa": { + "parameters_hash": "857e2f26da4895ea3d3f252a1cd4693d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:53:02.069560", + "end_time_simulation": 59.97999954223633 + }, + "4283f76dc4af4e819642d5e62b26f9a0": { + "parameters_hash": "41b40cf79972c1307da297c719b883a9", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:01:14.099082", + "end_time_simulation": 59.97999954223633 + }, + "07072cd293bd564dae2ed1074ae4b43d": { + "parameters_hash": "e2b49f7458ef25b74e1739a7dabc92d4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:01:16.710999", + "end_time_simulation": 59.97999954223633 + }, + "a062ad2ae4014ed089cfee7680910582": { + "parameters_hash": "a3259fd3d353ec22f28ebe7dad1959f4", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:01:19.784834", + "end_time_simulation": 59.97999954223633 + }, + "4eca4fe71aab5b04abdffc7e1868d5bb": { + "parameters_hash": "06f50f597e2d8df25129577c5ebf55ba", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:01:22.589003", + "end_time_simulation": 59.97999954223633 + }, + "8e89c852e0c451599249798ba65ca829": { + "parameters_hash": "b692be1033d692bc7bb0dd577ccd1df1", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:01:50.200385", + "end_time_simulation": 59.97999954223633 + }, + "9b3bb84455d355098220de0a837535bd": { + "parameters_hash": "d0655dd9b0e959d1a5ef4b798378f521", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:01:52.954168", + "end_time_simulation": 59.97999954223633 + }, + "72437358bbd7550cb4101e9a3688cfcb": { + "parameters_hash": "df6aa768a25a42f3e2e4a6b81bb0569d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:01:55.369946", + "end_time_simulation": 59.97999954223633 + }, + "270eef9d5f645b7ab32e4db525301172": { + "parameters_hash": "cea4a9ff781e70c98ede2f8cde61f6b0", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:01:58.282978", + "end_time_simulation": 59.97999954223633 + }, + "2a62aa2ef80c5d3e95eb002b9044e173": { + "parameters_hash": "8297e9449a3541477fc1ee355d1cc85d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:00.995982", + "end_time_simulation": 59.97999954223633 + }, + "67c56690e18a5b4cb20550ab3a7dc658": { + "parameters_hash": "5c7975d0003426c1ef936dbc83ee402c", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:02:03.756283", + "end_time_simulation": 59.97999954223633 + }, + "9aee7a23adfa5a74a9239e2c41d444e0": { + "parameters_hash": "8d76a40656ba73beb2229eccb7c136ad", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:06.208842", + "end_time_simulation": 59.97999954223633 + }, + "69baf539b70f519a98993dc9e1ea0285": { + "parameters_hash": "8a8fa2aacece15501bbf50668486a1c4", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:02:09.399308", + "end_time_simulation": 59.97999954223633 + }, + "9494cd24ca4c58aeb583db871eb80584": { + "parameters_hash": "c4bef4d89c9d52502e98719f7e80821c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:11.863151", + "end_time_simulation": 59.97999954223633 + }, + "ba3bddb0b2dc5ae7b3d73daf6246a0f8": { + "parameters_hash": "8a941619f131a3a5397a69b83a5517ed", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:02:14.756404", + "end_time_simulation": 59.97999954223633 + }, + "7b25a4650fae56e0ac3286cf16992b6d": { + "parameters_hash": "0a24982508684e3ea61ed44d361b6979", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:17.354438", + "end_time_simulation": 59.97999954223633 + }, + "c84f53da39df55daa3a857d18922e5b2": { + "parameters_hash": "d80f2243df053c5396e1a5cb6445a25f", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:02:19.974829", + "end_time_simulation": 59.97999954223633 + }, + "c6608ed52b09503fbec7b61a67d1e176": { + "parameters_hash": "6022c56460617aed258f4e15c9fe86d2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:22.551278", + "end_time_simulation": 59.97999954223633 + }, + "a519760d25ff5f5f9ef87bb4a5e09d4d": { + "parameters_hash": "2eceea3cf5aad8d8edec72d6c65499be", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:02:25.330704", + "end_time_simulation": 59.97999954223633 + }, + "54bcb25725a5574aa4ba7ecc9782662e": { + "parameters_hash": "efec0aacda48372ff5a2f3852e519dfd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:28.120376", + "end_time_simulation": 59.97999954223633 + }, + "12043ab374da5a9a81e6ebef609ef008": { + "parameters_hash": "e0da019d7599c6756d4ff71797629620", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:02:30.961313", + "end_time_simulation": 59.97999954223633 + }, + "40134052ae2d5b308968ab847a7851e3": { + "parameters_hash": "69b9bde33eadd55584ffb6c79ed04286", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:33.608218", + "end_time_simulation": 59.97999954223633 + }, + "f12e13839aa35418b517ce7c64e0da71": { + "parameters_hash": "c3e40838c98bf1691452f124d0d674ac", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:02:35.978310", + "end_time_simulation": 59.97999954223633 + }, + "4f60288e7a90582a841638ff2aaca5cc": { + "parameters_hash": "843f2e5668569c0c4c2a29931dc3b2a9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:02:38.637816", + "end_time_simulation": 59.97999954223633 + }, + "bc31ee1fb74c5e0785190e7f1140323e": { + "parameters_hash": "795ca5ab13741e67e762cdb7b7537926", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:41.245667", + "end_time_simulation": 59.97999954223633 + }, + "8a2750320f2855329efd768221edfb24": { + "parameters_hash": "03caffbe3cbd7107b9b4b509db08e5b5", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:02:44.382092", + "end_time_simulation": 59.97999954223633 + }, + "462d24e33046559da69626a148e64cd2": { + "parameters_hash": "ea53f06b904b2fed7700a6e0db12a8d9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:47.096278", + "end_time_simulation": 59.97999954223633 + }, + "ee03bc15507b5c3ea2eaa42824cb0b50": { + "parameters_hash": "fde77c19e66bc80a0ce060852ae542bf", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:02:50.217981", + "end_time_simulation": 59.97999954223633 + }, + "a10ff3094133540394ab079d1756763b": { + "parameters_hash": "1fe302022caf32c89e029f8fa408432f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:52.719459", + "end_time_simulation": 59.97999954223633 + }, + "3ff3d291e19c586a958adb1f94864c17": { + "parameters_hash": "c8e0fe58f23de5f3f75383e8f0ea9eee", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:02:55.530148", + "end_time_simulation": 59.97999954223633 + }, + "93404cca302c5425a0c3b262866ae0d6": { + "parameters_hash": "1eddd376ee4842c4dfea2a0675fb340d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:02:58.179428", + "end_time_simulation": 59.97999954223633 + }, + "44fba6b4ac86524b82c1a7f67897cfa6": { + "parameters_hash": "09683dbfa4badb201f46199ac036bb4c", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:03:01.198228", + "end_time_simulation": 59.97999954223633 + }, + "1077d622b1c653fca007e4f280485325": { + "parameters_hash": "0c6048996867acd653d004236b870a91", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:03.670489", + "end_time_simulation": 59.97999954223633 + }, + "062a37da2b3c523ea65fe50198329901": { + "parameters_hash": "e25f2297ed37ce231a36f6642322d064", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:03:06.284782", + "end_time_simulation": 59.97999954223633 + }, + "5833d0a994535b2cb60dd983cd38f327": { + "parameters_hash": "2d393f6c620279f022dfe164b05fe135", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:08.512007", + "end_time_simulation": 59.97999954223633 + }, + "c3b3ddcba6ff5526bd21d1d06ee66506": { + "parameters_hash": "e6d85212838c48e229c1ed574fb84361", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:03:11.497872", + "end_time_simulation": 59.97999954223633 + }, + "39b1a612c0c9567fb551fddead437b66": { + "parameters_hash": "a215f93822419214d94f4ce35af267fc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:14.259175", + "end_time_simulation": 59.97999954223633 + }, + "1e6654d17d454951848ea2e61337a186": { + "parameters_hash": "04d9dc6e6bff3d76299309521aae163a", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:53:04.540452", + "end_time_simulation": 59.97999954223633 + }, + "7fc4076866995d868789cdb925ed3e7e": { + "parameters_hash": "ea234d6c07e0c8388f067964bb3a49cc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:53:07.537666", + "end_time_simulation": 59.97999954223633 + }, + "00a3bc77dae04714babe89a17c0a3c7e": { + "parameters_hash": "f5f25c9349c0b7f30682284831ecf183", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:53:11.068745", + "end_time_simulation": 59.97999954223633 + }, + "c5fff46b76df59b6be3e53eed2d13c91": { + "parameters_hash": "4fd2c61654caf77e72d4562d354f5fae", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:53:13.884496", + "end_time_simulation": 59.97999954223633 + }, + "c131520e5f80470ebb16965a0c6ba8f4": { + "parameters_hash": "829452e9de1c781c9cd4e755f45b1cdb", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:01:38.752816", + "end_time_simulation": 59.97999954223633 + }, + "23e46579996056c4bcb62501e1a891d0": { + "parameters_hash": "6a4573864275c36882628125c92781dd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:01:42.339183", + "end_time_simulation": 59.97999954223633 + }, + "8549e70b7f975ae68b059707d6673904": { + "parameters_hash": "5f3821630bc45ce5aa5333059a160167", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:03:16.735683", + "end_time_simulation": 59.97999954223633 + }, + "dc9aad2835b05b70bcd8eb896031d295": { + "parameters_hash": "815cdf42ea10486072e7c23256ca8b78", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:03:19.620213", + "end_time_simulation": 59.97999954223633 + }, + "08ac616e219a592c8751b456437ae83a": { + "parameters_hash": "8f92bb47e4bc9e0eda759e38a34f6dc6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:22.459806", + "end_time_simulation": 59.97999954223633 + }, + "def64a518bb95ea3a3417e479e9c18cd": { + "parameters_hash": "ec5fe8acdd68c3db312ab706e8637708", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:03:25.637251", + "end_time_simulation": 59.97999954223633 + }, + "62136e351234555d934ea6ff3ce42c93": { + "parameters_hash": "d5ad99367d92d20bd17c83e32e6056e9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:28.092242", + "end_time_simulation": 59.97999954223633 + }, + "5af136001c8e55ed96639fb169b820df": { + "parameters_hash": "b339372c4d697c6866bfa02ddb77fd53", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:03:31.047763", + "end_time_simulation": 59.97999954223633 + }, + "e2808a3827805dea8606d761ca2fa0b1": { + "parameters_hash": "56dd5ae69a1fb9e2d6d37f8b5b5a0198", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:33.610272", + "end_time_simulation": 59.97999954223633 + }, + "a13379a350805d2da9c1dcac35e69f94": { + "parameters_hash": "7cc35d3d5260480e56af8ebe0db7d7b4", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:03:36.081081", + "end_time_simulation": 59.97999954223633 + }, + "26eb8e4337bc5ec4825c5dc396f0b19d": { + "parameters_hash": "32739aaede06793fd9acbad485f1d0be", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:38.522448", + "end_time_simulation": 59.97999954223633 + }, + "5849be8c04945d069415611631fa89df": { + "parameters_hash": "a3bb18e7659b8242c9cf0bb1a20546cd", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:03:41.227195", + "end_time_simulation": 59.97999954223633 + }, + "c98c92172e195d0fb89534843d86d792": { + "parameters_hash": "d58d6ec22e4b854eab551163ed1bce6a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:44.053857", + "end_time_simulation": 59.97999954223633 + }, + "bd67c3ae184956f197d8d52b5b0934f1": { + "parameters_hash": "0e1024e326eff52d718d161cfb67c31c", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:03:46.946046", + "end_time_simulation": 59.97999954223633 + }, + "89dbe52ddf355d59b40c99bdef041f00": { + "parameters_hash": "9e24792ae82ae5d60196f1fc32f2774c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:49.758438", + "end_time_simulation": 59.97999954223633 + }, + "3f14e6ab62885cecad3876e7b3caf6dd": { + "parameters_hash": "f6112bf6818c85259ec05f4aef8013d6", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:03:52.511535", + "end_time_simulation": 59.97999954223633 + }, + "c03ccb145355525b91a80025bdd4fb65": { + "parameters_hash": "0184c2279b09e0f286c8f459e0f165b4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:03:55.435785", + "end_time_simulation": 59.97999954223633 + }, + "e4cf1fe8d4165cb1845b624096980de5": { + "parameters_hash": "149c92837125f8dc1559cabfcc6f11d2", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:03:58.500811", + "end_time_simulation": 59.97999954223633 + }, + "9458101d2ecd5bf884b3070f68c69a71": { + "parameters_hash": "0dc4b7b3f02ab54bcf6749f7c38d2b41", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:00.991670", + "end_time_simulation": 59.97999954223633 + }, + "e4f1751964274f91b0d37343a02d8a4b": { + "parameters_hash": "48c8c2739152d228491e337aa7aca2dd", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:53:17.111212", + "end_time_simulation": 59.97999954223633 + }, + "b52ea2bad4015d5888267c5dabc0ccb1": { + "parameters_hash": "d8eb5e7cd977849b25aee8aec093bb94", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:53:20.270374", + "end_time_simulation": 59.97999954223633 + }, + "1276239e6f2a43c58fe7ab267bd41d94": { + "parameters_hash": "7a563abc0a8b48d95b01f5708a0c44cd", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T23:53:23.336021", + "end_time_simulation": 59.97999954223633 + }, + "6c46acbf4e0a5596a295c9ad22f4ace7": { + "parameters_hash": "cf5ea79c4f1fe4cdf2c7c8097306813a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T23:53:26.020375", + "end_time_simulation": 59.97999954223633 + }, + "9820ec472b7e4371a4d40be054b4c6ae": { + "parameters_hash": "8b8b3520edb16081225b5692f78e7127", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-06T00:01:56.988763", + "end_time_simulation": 59.97999954223633 + }, + "7a246520055858f2bb24f3c881b45653": { + "parameters_hash": "4ad52d003d88c64b98809ce7e5d76d8c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:02:00.107553", + "end_time_simulation": 59.97999954223633 + }, + "1530c3994eb847d0a9a66fe1d6584de6": { + "parameters_hash": "0d1a71336672ed0528f5d34bfeb4ed3d", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:02:03.338682", + "end_time_simulation": 59.97999954223633 + }, + "c271c73a35f755399d57752347699c62": { + "parameters_hash": "7a362aa244b66d92bbd2e803a1b92a22", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:02:06.269796", + "end_time_simulation": 59.97999954223633 + }, + "38723bf741f84bc89f2f31302035b4f2": { + "parameters_hash": "89a8fd37bef8de680fc6bba433cdb244", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:02:09.304213", + "end_time_simulation": 59.97999954223633 + }, + "9bf9968b6dbf5a44b1045d2403d816c7": { + "parameters_hash": "9e03e6b0b0fbaeea060f5af4eccc2786", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:02:12.491441", + "end_time_simulation": 59.97999954223633 + }, + "2c5b25ed588c52f5a5b4f58fc232e4bb": { + "parameters_hash": "6fe2b7c42d4485213cf19077ee205163", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:04:03.971011", + "end_time_simulation": 59.97999954223633 + }, + "95ed9fef81a25479881312e047965573": { + "parameters_hash": "50f704f5c9b607dded1292f4f7629096", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:04:06.706830", + "end_time_simulation": 59.97999954223633 + }, + "5f4743225ebc5c65bffe1724dc7d8050": { + "parameters_hash": "cbe8cf14d4e12647fc4dd383322051ee", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:09.099332", + "end_time_simulation": 59.97999954223633 + }, + "a9876f8562df54ab9ce89706f944084b": { + "parameters_hash": "c601f5504325c19412f83f8af6c4ea2f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:04:11.700429", + "end_time_simulation": 59.97999954223633 + }, + "a71c54886b9552d8b931c513674216c5": { + "parameters_hash": "147095b26582bd2ffdd2f2fd49622cdf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:14.432020", + "end_time_simulation": 59.97999954223633 + }, + "104f1a820a9c5771a92dd0511493c1da": { + "parameters_hash": "bc3296c90d8b4a7bef8fe35b4d434bce", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:04:17.253714", + "end_time_simulation": 59.97999954223633 + }, + "26780ada23ed56fbb48ddabdfbe88c4a": { + "parameters_hash": "cf4f7493a102421851e4d30caf8639ed", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:19.848935", + "end_time_simulation": 59.97999954223633 + }, + "a00033198f2559f3b1f712a582a9e963": { + "parameters_hash": "3d98d8803f852d21795b13a78a207cc2", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:04:22.805110", + "end_time_simulation": 59.97999954223633 + }, + "504346a4edd150f49a17575946fcd657": { + "parameters_hash": "4a65f3e6e24051c4f0cce7ae0643b1e9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:25.518303", + "end_time_simulation": 59.97999954223633 + }, + "e30d2d00d5d45d71be566d22fabbba7b": { + "parameters_hash": "d1e2fdf33846245c9957e6e3541ba774", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:04:28.485545", + "end_time_simulation": 59.97999954223633 + }, + "6936dda9b3ab5167a5452e562debdbd3": { + "parameters_hash": "5c3e4909ce4abd14e30f8d68fc7cc3d4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:31.741274", + "end_time_simulation": 59.97999954223633 + }, + "fb36caf39e8c5f1299b26503ec006cb7": { + "parameters_hash": "01fda366414dd14b8d17f3c7d76cadbd", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T22:19:04.342060", + "end_time_simulation": 59.97999954223633 + }, + "1e759179ee4a5a61ba29277ae684db7b": { + "parameters_hash": "e1d9da7431e39d66ef10db0b86acb3bb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T22:19:07.738145", + "end_time_simulation": 59.97999954223633 + }, + "3b1549f0f19c516a83a37612f35671cb": { + "parameters_hash": "8b83020e9c34bef199471a0a026550d3", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:04:40.421197", + "end_time_simulation": 59.97999954223633 + }, + "9bbecd35f2cd570eaac554703d9118be": { + "parameters_hash": "c79bd1f28b0e01bdd374f5827191db12", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:42.954296", + "end_time_simulation": 59.97999954223633 + }, + "682ebbf9a88354c5874809941c1fba8f": { + "parameters_hash": "41c394f48997a0ac54cfa950afb23e5b", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:04:45.959719", + "end_time_simulation": 59.97999954223633 + }, + "6e0d0e82ddee538baf0cbfcfe2a5de51": { + "parameters_hash": "6275c146365d4d1b5f8f8fc85ac81248", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:49.086968", + "end_time_simulation": 59.97999954223633 + }, + "8405a426cabe594ab796553c977bef1e": { + "parameters_hash": "e9ed1afd8b5f1404a139536056e6e650", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:04:52.527390", + "end_time_simulation": 59.97999954223633 + }, + "a72555f175cc5e8aa302634ba9bb6918": { + "parameters_hash": "cc060624dfa817dbc69d7f18b9957d5c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:04:56.381221", + "end_time_simulation": 59.97999954223633 + }, + "1bc4f20c97f85f32894348c03a82530e": { + "parameters_hash": "b897ee6a359bd92f879d9d70d34da7b4", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:04:59.682090", + "end_time_simulation": 59.97999954223633 + }, + "5cee759d82fc55aea1483e448767b37c": { + "parameters_hash": "909e1eb68077a0c160d6540080fecb8b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:02.594392", + "end_time_simulation": 59.97999954223633 + }, + "68c17889e0f146f7a0772ecd1c0e63cf": { + "parameters_hash": "3681ea0983bf36d9744aa3516845b1b1", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:02:33.107148", + "end_time_simulation": 59.97999954223633 + }, + "9e609951e2cf5b559e88031197e96a00": { + "parameters_hash": "afc29bde06a75fa992e45ae8295af8fb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:02:36.275519", + "end_time_simulation": 59.97999954223633 + }, + "0dda565c44d35ddc944cae93e806df1d": { + "parameters_hash": "698196ffa010965e77369f23d25dee13", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:05:05.361401", + "end_time_simulation": 59.97999954223633 + }, + "5ed1c131e780539b84a3e3c3e76d73d4": { + "parameters_hash": "c37c7d430d6417a488dc85bcf8acc26b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:05:08.842321", + "end_time_simulation": 59.97999954223633 + }, + "5776de3eb4405d169b7cbf727763fc3c": { + "parameters_hash": "c951309a0ed643f86c92809fb61d68a9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:11.612085", + "end_time_simulation": 59.97999954223633 + }, + "a3f71a6fa2d753678c095a7c149226c2": { + "parameters_hash": "cc6243ab9019f61a9863cc6fa16cd4c6", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:05:14.538995", + "end_time_simulation": 59.97999954223633 + }, + "e323d236ccaf549f9389fa97b1434301": { + "parameters_hash": "bdbc11f7bc29a80f4a7f550b65b95c7e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:17.530680", + "end_time_simulation": 59.97999954223633 + }, + "ba1a51d50f0b5031b0f10f2b48f8f5b5": { + "parameters_hash": "8eec429a21e528fd576f160066e454dc", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:05:20.705411", + "end_time_simulation": 59.97999954223633 + }, + "6c7260009d2656e69a24b8bb5f10ef46": { + "parameters_hash": "fd6585afa998cf4a0bca312789850c4c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:23.392377", + "end_time_simulation": 59.97999954223633 + }, + "45aa02ea71c0510aae5f198aef30fef5": { + "parameters_hash": "680349aa8ffea35b1f3bc0174b398a0d", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:05:27.231962", + "end_time_simulation": 59.97999954223633 + }, + "aae91fbbb9995f7ea9cfb49d3646ac0f": { + "parameters_hash": "ee0bfaf051aab1ab07ec0fe631ea13ff", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:30.667112", + "end_time_simulation": 59.97999954223633 + }, + "1b8aec0c38925e06825cf0c9935bb7c5": { + "parameters_hash": "c5bd39968957c0515b22669d8a18c145", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:05:34.206822", + "end_time_simulation": 59.97999954223633 + }, + "7564c899df125067bb975cf94e344115": { + "parameters_hash": "aebf8bb37569600bd91f36e6f8e16fbd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:37.339613", + "end_time_simulation": 59.97999954223633 + }, + "08103a7800b856039bade05a970ca1ee": { + "parameters_hash": "01a10f69b5821bd75a3a20c157c61f49", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:05:40.306353", + "end_time_simulation": 59.97999954223633 + }, + "072eb68a72f15915b48638e5882fcb2c": { + "parameters_hash": "b7717cd08e66b927e5e9e767c1692c51", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:43.211638", + "end_time_simulation": 59.97999954223633 + }, + "64b39ac8f9895a7697d242b7e269419c": { + "parameters_hash": "2ca6c469978369834cee2ecd18e815a3", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:05:46.617111", + "end_time_simulation": 59.97999954223633 + }, + "e03eab6afaa855b3a8b6d9c599a990aa": { + "parameters_hash": "7c2be0e199a32ef168369ec95f88a807", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:49.859498", + "end_time_simulation": 59.97999954223633 + }, + "209a3c1fdd0f583d95cb0d1c0ebc522d": { + "parameters_hash": "2d37bd60695c2c4eae7d2c7e2e8ac599", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:05:52.789891", + "end_time_simulation": 59.97999954223633 + }, + "3be9112bc00f58a9adefe50b1b2854bf": { + "parameters_hash": "170a3766566cdce499f42c0b7fdf78c7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:05:56.110059", + "end_time_simulation": 59.97999954223633 + }, + "42a04eb48f2359898f15b1581dab6727": { + "parameters_hash": "09d201c2304591700da298bc6e543570", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:05:59.520024", + "end_time_simulation": 59.97999954223633 + }, + "ca297a9088665b8481f1b37d5b764637": { + "parameters_hash": "32ad6492b807548ede9d495c1cfde2bd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:02.734760", + "end_time_simulation": 59.97999954223633 + }, + "6ca1f93c5f9d59669a10db584518a5da": { + "parameters_hash": "5598a5fe92b467882d3a4e316c2ac107", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:06:06.112651", + "end_time_simulation": 59.97999954223633 + }, + "b4d9b743ae235dad9d6a4db019115179": { + "parameters_hash": "1809b4ca4840e37e7864d901a88f5a0f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:09.016924", + "end_time_simulation": 59.97999954223633 + }, + "c397050fcfde4360acb33eb8acbc71f7": { + "parameters_hash": "b27ece52b4887881cd09e39efb79a3cd", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:02:52.562399", + "end_time_simulation": 59.97999954223633 + }, + "7e3cdaaf2233543aa4953d6ee62c3120": { + "parameters_hash": "3356478d04795d99acc35fbef88d8bf5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:02:55.387820", + "end_time_simulation": 59.97999954223633 + }, + "cefd2b01da504c848b1a81d58c9ef1a2": { + "parameters_hash": "ff1c1f9029d44b6389ae14bfee4d83e5", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:02:58.238010", + "end_time_simulation": 59.97999954223633 + }, + "2dd1a254a4745309837ff3668c14a6a2": { + "parameters_hash": "466f75b858ba165e08ce0af8df5b01af", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:03:01.216629", + "end_time_simulation": 59.97999954223633 + }, + "e9cc36d951655bdfbcf90d0bb2322a0e": { + "parameters_hash": "559649d78778f524144fd61b98183086", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:06:11.661426", + "end_time_simulation": 59.97999954223633 + }, + "90872800e402540eb77c39d26a123b23": { + "parameters_hash": "fda06dc08847de9fb0c4a88b7350dfc7", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:06:14.708796", + "end_time_simulation": 59.97999954223633 + }, + "4028376c2dac53c5a6443457100463d1": { + "parameters_hash": "05ce6b45de23423ef8c380b0286634c0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:17.440437", + "end_time_simulation": 59.97999954223633 + }, + "9698d8207595571e9187650324239bb0": { + "parameters_hash": "c554c189cfd1f44941d4ad41c2d8e289", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:06:20.287193", + "end_time_simulation": 59.97999954223633 + }, + "5dd709fd21085057902dffb95bad7571": { + "parameters_hash": "b333eeeece8aa3da23c56fa97b217fb7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:22.671389", + "end_time_simulation": 59.97999954223633 + }, + "be961034c82352e09e4c5952be16637a": { + "parameters_hash": "5e6f000fa3912ff2d686d24bef62f9b6", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:06:26.035385", + "end_time_simulation": 59.97999954223633 + }, + "5f0759cdfe2a54459b5295bbcc6d3be2": { + "parameters_hash": "66c7edadf10b38378d8bdb599eab2258", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:28.806129", + "end_time_simulation": 59.97999954223633 + }, + "6be5dc20d4365c68ae8ef3bae8a17615": { + "parameters_hash": "75b4ec8c7754d1815c1a0a1c4566827a", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:06:31.744255", + "end_time_simulation": 59.97999954223633 + }, + "5c5218c3b6065cc99f943deb676e8937": { + "parameters_hash": "f63efb58748d666c42745a292fcf593c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:34.402018", + "end_time_simulation": 59.97999954223633 + }, + "1de56b2f3fe85f3caabdf92a064abfbd": { + "parameters_hash": "60b020eec80d9c642b5a7709740a5421", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:06:37.616842", + "end_time_simulation": 59.97999954223633 + }, + "a727f1b17927574bb0077e7c5cfa91c8": { + "parameters_hash": "1121f921c8d7edea7612bc09241edc0d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:40.232643", + "end_time_simulation": 59.97999954223633 + }, + "96f1621029835fd49654f3249386c245": { + "parameters_hash": "5d776197ac4eab709532b359e5c94428", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:06:43.398189", + "end_time_simulation": 59.97999954223633 + }, + "37b4cfe71e9b53c392948d4b0e065f35": { + "parameters_hash": "56b557026dc143c94540612394dd1fd3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:46.027704", + "end_time_simulation": 59.97999954223633 + }, + "2e254fceb17757d7b6ab348723911676": { + "parameters_hash": "682083c2fb29ff32fdb4ad3c9c35dcf1", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:06:48.918701", + "end_time_simulation": 59.97999954223633 + }, + "c292ebf0bb9e55869c34dfd7e5a2488d": { + "parameters_hash": "b3e97460de236aa359b01d762a8ec4db", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:51.345614", + "end_time_simulation": 59.97999954223633 + }, + "827882d3769753ea833aa5fb134d3c48": { + "parameters_hash": "01d8e6bd4a2cf33787535caef68e273b", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:06:54.233933", + "end_time_simulation": 59.97999954223633 + }, + "6bf6337b72ac57f486d0f019e2fe107d": { + "parameters_hash": "65287e23070460e95178e4913227e382", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:06:56.617244", + "end_time_simulation": 59.97999954223633 + }, + "fdcb330163885a46a5789d67993122f1": { + "parameters_hash": "ec204acf2d840d2371d33c83d37ea845", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:06:59.307829", + "end_time_simulation": 59.97999954223633 + }, + "ab8acd3496c157df93a629ec3a40df8e": { + "parameters_hash": "29c0a33419c282d86086dbf8954fe7db", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:01.498660", + "end_time_simulation": 59.97999954223633 + }, + "e6614388019a4d22bdd9140403a47547": { + "parameters_hash": "0f603778fccc9f44ee65fb6bf8140069", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-06T00:03:04.205693", + "end_time_simulation": 59.97999954223633 + }, + "c46e8d0f76d15e87a9b0dce0dcd84776": { + "parameters_hash": "fff22e4395e87bf297288e95c12c74df", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:03:07.436170", + "end_time_simulation": 59.97999954223633 + }, + "7379873992335258a1c5571e0f7902db": { + "parameters_hash": "1f1a2fb77d47733588a417e23660ac4f", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:07:04.016965", + "end_time_simulation": 59.97999954223633 + }, + "e755a27e73ee5af2bf1337a1a71d0ea3": { + "parameters_hash": "21262e359a3b624c46be559d7db5aae5", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:07:07.052811", + "end_time_simulation": 59.97999954223633 + }, + "2cb73bf455f9592f84239d88b9fe3fe9": { + "parameters_hash": "abb847064b7eeb833ccaeed1be4e53cf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:09.801332", + "end_time_simulation": 59.97999954223633 + }, + "915a6d979e4551aa9098480bf6c2269f": { + "parameters_hash": "b952f415872c0a08fe63b3a9bc3918d9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:07:12.753979", + "end_time_simulation": 59.97999954223633 + }, + "f7d32cf150b65da386fffd4d0e88fd3a": { + "parameters_hash": "db6e6b6c979d4b2d70e621197e0b23af", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:15.490286", + "end_time_simulation": 59.97999954223633 + }, + "7641ef2c46bb5e83953c8dbd5a1efe6e": { + "parameters_hash": "0f83baf6e2f6091c4f1997f0ec09b2fd", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:07:18.527278", + "end_time_simulation": 59.97999954223633 + }, + "32794a9e074956998c5b2e3dedf3acae": { + "parameters_hash": "2cc11832b0b34830f0842d350ad1ff48", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:21.051879", + "end_time_simulation": 59.97999954223633 + }, + "fc653cb5b3c05a259ddecd3547667a15": { + "parameters_hash": "9855205bcbf0bf2f7274ab7b17974d8e", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:07:23.936596", + "end_time_simulation": 59.97999954223633 + }, + "b1c072a798455b279fd6c5370d110019": { + "parameters_hash": "9eab2293328ff3563142a97933b9de31", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:26.678755", + "end_time_simulation": 59.97999954223633 + }, + "b12f6205eb85528db5ebf292514a137b": { + "parameters_hash": "5355b5ab2539556c90666a9dc08093fa", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:07:29.890225", + "end_time_simulation": 59.97999954223633 + }, + "bc10592530595ffc8b7898a331efe856": { + "parameters_hash": "d71ab19dde37393536827797bfd88b9c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:32.589696", + "end_time_simulation": 59.97999954223633 + }, + "90573cfec59c56f7a6310cd9d601c0c5": { + "parameters_hash": "cf0618f0970b0cba4c1cc2e30e6fd2cc", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:07:35.100423", + "end_time_simulation": 59.97999954223633 + }, + "1e933fa8994c5c54849b3d19b16edcf7": { + "parameters_hash": "ebe36772714d1b36c16fb2568c4d6900", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:37.528824", + "end_time_simulation": 59.97999954223633 + }, + "21a2d71e900a5028b7c01a97f809536a": { + "parameters_hash": "5564e9fa9ec19bd985779b1fe7b9f335", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:07:40.582763", + "end_time_simulation": 59.97999954223633 + }, + "2ed92f63664b547584529740086fd06c": { + "parameters_hash": "11d8512d48ac8bc43180b60e887f6d4a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:43.359843", + "end_time_simulation": 59.97999954223633 + }, + "b0153a64f01750daa5a495504f85005f": { + "parameters_hash": "67125afed4e08655a525c6657146c83c", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:07:46.255090", + "end_time_simulation": 59.97999954223633 + }, + "21eb5eaf2bc855c395dac49717b071b8": { + "parameters_hash": "41ccb8ee28a03f81c0dac670f75d6b10", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:49.127263", + "end_time_simulation": 59.97999954223633 + }, + "aaaa2d3efcad5dd1a58b6a7616798eca": { + "parameters_hash": "a7dc0d87d0281220d9ca7aeb262b2866", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:07:51.889139", + "end_time_simulation": 59.97999954223633 + }, + "b8457283e6f05b2381310d4e664ca179": { + "parameters_hash": "73cf9619d9466479b14569a451a11a26", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:07:54.353668", + "end_time_simulation": 59.97999954223633 + }, + "cc34c415d1465097ae993175b76900b8": { + "parameters_hash": "a29268334ba4a9303e4362444dee0979", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:07:57.742135", + "end_time_simulation": 59.97999954223633 + }, + "73f2e78829c353489f629b030fb60647": { + "parameters_hash": "28bfb7b56e77743610400dd748643e52", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:00.542642", + "end_time_simulation": 59.97999954223633 + }, + "b248181b23624fdabde51fb023f87d70": { + "parameters_hash": "bc02b76c7cd403f0fca3090e321f8414", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-06T00:03:40.620016", + "end_time_simulation": 59.97999954223633 + }, + "bdc4bfc0300d5a6bbd2b8c07b6cbdcec": { + "parameters_hash": "4364d559289d8bdd3d2ba19aa31d1fec", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:03:43.507329", + "end_time_simulation": 59.97999954223633 + }, + "3828621236495616b794ce84cf8c84ea": { + "parameters_hash": "646d80e73dfb4473643ec711415ec877", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:08:02.923982", + "end_time_simulation": 59.97999954223633 + }, + "acd65eb4f78a5539a10bb7f268c00b8b": { + "parameters_hash": "aa813762880b0c48ade181b6ffc371fb", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:08:05.631721", + "end_time_simulation": 59.97999954223633 + }, + "7e3b590f83ce550bad1dafef11d48f46": { + "parameters_hash": "518fa36ac3cf409ed089e9e71e940e3d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:07.883323", + "end_time_simulation": 59.97999954223633 + }, + "2736666ef08051b18037717bad48edcc": { + "parameters_hash": "9c9d519dabd56d172e09e05754193d47", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:08:10.826384", + "end_time_simulation": 59.97999954223633 + }, + "c1cc301fe297502e9457785e7d926903": { + "parameters_hash": "de65f7819de40e71ba0ba5ffe1a095e7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:13.513721", + "end_time_simulation": 59.97999954223633 + }, + "b4cfcd7a4a065adca4415e60bc9978be": { + "parameters_hash": "4f5604abc94855b7f6cf3375d1115771", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:08:16.490447", + "end_time_simulation": 59.97999954223633 + }, + "5ccebda3a95d57ef9b2c0f9cf8c741ac": { + "parameters_hash": "d36d1a0769b674af61dea9269760392b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:19.114740", + "end_time_simulation": 59.97999954223633 + }, + "6407da2f934a51c5951cb7161ace6674": { + "parameters_hash": "71151601bc29c119935779b5ae768f14", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:08:22.185492", + "end_time_simulation": 59.97999954223633 + }, + "c2500b97bb415439a9bc82d9343c2f09": { + "parameters_hash": "7c2df4d1bf0dfd75b344eef4860eaac5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:25.024194", + "end_time_simulation": 59.97999954223633 + }, + "f0995aff99ff5f638e78ffcd4af03f11": { + "parameters_hash": "773a09acd015788cf8e103e4ec611bce", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:08:28.215301", + "end_time_simulation": 59.97999954223633 + }, + "cfc3f7b12ff858e3b1f987876d6f41f9": { + "parameters_hash": "d76738586cb642f7d3f5191a04246c89", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:30.824332", + "end_time_simulation": 59.97999954223633 + }, + "b2728ee12aa05150b94e4b1938af6e7f": { + "parameters_hash": "95fe793bb6fa720df73a48d757316870", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:08:33.571235", + "end_time_simulation": 59.97999954223633 + }, + "a12acbe2cdbd5d4e80932131576b3c95": { + "parameters_hash": "3ada9060b6dedcb87e76669162c258ca", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:36.452226", + "end_time_simulation": 59.97999954223633 + }, + "780baff27b6656a6905e38085f35a5eb": { + "parameters_hash": "977613f8bbbce8857da9c5104eb9b142", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:08:39.103337", + "end_time_simulation": 59.97999954223633 + }, + "c79822d169f751e0b99adc1320c96ba6": { + "parameters_hash": "1a47eb7da250afe7c9ffd7f22ab6a059", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:41.642574", + "end_time_simulation": 59.97999954223633 + }, + "0310deb38f465db18ad47c098129e23d": { + "parameters_hash": "b443c971956232fc19386c6fc839e29f", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:08:44.566155", + "end_time_simulation": 59.97999954223633 + }, + "539e29fd24fb5cffaa4fdcbcd2d2eefa": { + "parameters_hash": "36deddbd14d7466769ddf4dcbbcc1506", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:47.252970", + "end_time_simulation": 59.97999954223633 + }, + "461103bd3ac255b2a1e6bd5a4122cbc4": { + "parameters_hash": "1998992a9f55b8529d050692671866b4", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:08:49.934093", + "end_time_simulation": 59.97999954223633 + }, + "4ce049f71c1059cfab0ecfa8355d0780": { + "parameters_hash": "c210baeab2ddd45e5cc3636b5dabd782", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:08:52.637603", + "end_time_simulation": 59.97999954223633 + }, + "551dcaf16f2646198becdde681de62ef": { + "parameters_hash": "e7da5eda77a0f1e4a81ddda03fdfeb57", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-06T00:04:11.452613", + "end_time_simulation": 59.97999954223633 + }, + "717adfe6735752269eebd6efb3dd39b7": { + "parameters_hash": "b991efbae63fdde2f4c23bb00481143e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-06T00:04:13.890202", + "end_time_simulation": 59.97999954223633 + }, + "3ce3324912e859abbad3c95979f6c0a8": { + "parameters_hash": "9b6eda6f9cb7815d76682d83fc00dc4c", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-05T20:08:55.023805", + "end_time_simulation": 59.97999954223633 + }, + "d3603aad0464539792747de81d5aeeb8": { + "parameters_hash": "98ad1cf1b278ec17e28e439c9fa5c282", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:08:58.049604", + "end_time_simulation": 59.97999954223633 + }, + "917e491609425dbabbb12dbbd03503f6": { + "parameters_hash": "8a6555e1e4f24dafbea31a3f36a25b6a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:00.793112", + "end_time_simulation": 59.97999954223633 + }, + "896836139b7b5d188e74e95be86f42e9": { + "parameters_hash": "601fdf04c1ee5717c3df7dfc67324f58", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-05T20:09:03.544300", + "end_time_simulation": 59.97999954223633 + }, + "193ad266eba752feaf78558e4b867795": { + "parameters_hash": "65f362cb4c37f026808e958e671705fb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:06.036460", + "end_time_simulation": 59.97999954223633 + }, + "3c789ac43d2352d8aa6bd93b83457cb6": { + "parameters_hash": "33c7b3cb1583a7e141577035b1a71a16", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:09:08.741730", + "end_time_simulation": 59.97999954223633 + }, + "530a6069fa4950f2b2c652e8ba7ede24": { + "parameters_hash": "a987bafa3bbc235af384966a2b0b3a9b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:11.228860", + "end_time_simulation": 59.97999954223633 + }, + "39568986f7f952c58d8a28ce42dad54f": { + "parameters_hash": "c1acf69a92d36c758cd506a3c394daad", + "run_type": "intervention", + "class_internal": "viscous_damping", + "class_agent_facing_name": "viscous_damping", + "status": "success", + "timestamp": "2026-05-05T20:09:14.213301", + "end_time_simulation": 59.97999954223633 + }, + "8148fc3a772b5e65b3abde90ecbefc6d": { + "parameters_hash": "90ec8cdb1bdd27b193684598dc9e5e90", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:17.116344", + "end_time_simulation": 59.97999954223633 + }, + "438eb83368255bf4b719102d3e79d605": { + "parameters_hash": "ec6a21de72937ec26ffbc4eeb85555bc", + "run_type": "intervention", + "class_internal": "restitution_right", + "class_agent_facing_name": "restitution_right", + "status": "success", + "timestamp": "2026-05-05T20:09:19.896924", + "end_time_simulation": 59.97999954223633 + }, + "4caf3d51c74859b68778c0e702ad1c9f": { + "parameters_hash": "333a71d8e0b5dbc5adb096ca6528dec4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:22.436119", + "end_time_simulation": 59.97999954223633 + }, + "fefe43b61bf75c928a2124106f48417b": { + "parameters_hash": "6045f68d0cc81361828cc4328c39b904", + "run_type": "intervention", + "class_internal": "restitution_left", + "class_agent_facing_name": "restitution_left", + "status": "success", + "timestamp": "2026-05-05T20:09:25.347663", + "end_time_simulation": 59.97999954223633 + }, + "ab1ec64876e75c2cb884ab35d6531567": { + "parameters_hash": "e0274626208ab4094e6083aa92a8c50b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:28.159725", + "end_time_simulation": 59.97999954223633 + }, + "58c3640b5dc9550f86a0e1b9da9a7a0d": { + "parameters_hash": "6c9bb117b10ec8c8a67658e98fe508ab", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:09:30.766624", + "end_time_simulation": 59.97999954223633 + }, + "a8517a4e17135b3ebc1750d95fc94d9a": { + "parameters_hash": "48e37ab9b7fc7917fcc91b7daf0a1f50", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:33.691194", + "end_time_simulation": 59.97999954223633 + }, + "f93bb9ba46dc5d02a3eff4a53fb205b7": { + "parameters_hash": "32cd1a0b13d1a9b5b731d990cd9ec120", + "run_type": "intervention", + "class_internal": "inclination_angle", + "class_agent_facing_name": "inclination_angle", + "status": "success", + "timestamp": "2026-05-05T20:09:36.583662", + "end_time_simulation": 59.97999954223633 + }, + "e711968a70f050b9afe0018cc3d40803": { + "parameters_hash": "37a95389e9edc10c85119b6cf10327fe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T20:09:39.308954", + "end_time_simulation": 59.97999954223633 + } +} \ No newline at end of file diff --git a/questions/BounceBall/noise_adder.py b/questions/BounceBall/noise_adder.py new file mode 100644 index 0000000000000000000000000000000000000000..16371e5711b5a77f011b5c910965637180c2b9dc --- /dev/null +++ b/questions/BounceBall/noise_adder.py @@ -0,0 +1,274 @@ +from __future__ import annotations + +import numpy as np +import pandas as pd + +from shared.noise_analysis import ( + DOCUMENTED_LOCAL_NOISE_ANALYSIS_POST_ROWS, + DOCUMENTED_LOCAL_NOISE_ANALYSIS_PRE_ROWS, + first_detectable_time_from_baseline, + quantify_analysis, +) +from shared.noise_snr import hash_string + + +_POSITION_COLUMN = "Initial_Spacer_F_x" +_SPEED_COLUMN = "Initial_Spacer_F_v" +_FORCE_COLUMN = "Right_Hard_Stop_f" +LOW = { + "position_base_sigma_scale": 0.0175, + "position_drift_sigma_scale": 0.004, + "position_event_sigma_scale": 0.04, + "position_quant_step_scale": 0.0025, + "position_quant_step_floor": 0.0005, + "speed_base_sigma_scale": 0.00088, + "speed_hetero_sigma_scale": 0.0022, + "speed_drift_sigma_scale": 0.0001056, + "speed_event_sigma_scale": 0.00704, + "speed_post_event_sigma_scale": 0.00308, + "right_force_base_sigma_scale": 0.0001, + "right_force_event_sigma_scale": 0.004, +} +HIGH_SCALE = 4.0 +HIGH = {key: float(value) * HIGH_SCALE for key, value in LOW.items()} + + +NOISE_DICT = {"low": LOW, "high": HIGH} +SNR_THR_DICT = { + "low": {"global": [-1.0e9, -1.0e9, -1.0e9], "local": [-1.0e9, -1.0e9, -1.0e9]}, + "high": {"global": [-1.0e9, -1.0e9, -1.0e9], "local": [-1.0e9, -1.0e9, -1.0e9]}, +} +_MAX_NOISE_RESAMPLE_ATTEMPTS = 25 + + +def _analysis_meets_thresholds( + noise_analysis: dict[str, list[float | str | None]], + *, + noise_level: str, +) -> bool: + thresholds = SNR_THR_DICT[noise_level] + for scope in ("global", "local"): + values = noise_analysis.get(scope, []) + limit_values = thresholds.get(scope, []) + for idx, raw_value in enumerate(values): + if raw_value is None or idx >= len(limit_values): + continue + value = float(raw_value) + if np.isfinite(value) and value < float(limit_values[idx]): + return False + return True + + +def _rng(seed: int, key: str) -> np.random.Generator: + derived = (int(seed) ^ hash_string(key)) & 0xFFFFFFFF + return np.random.default_rng(derived) + + +def _values(df: pd.DataFrame, column: str) -> np.ndarray: + return pd.to_numeric(df[column], errors="coerce").to_numpy(dtype=float) + + +def _finite_scale(values: np.ndarray) -> float: + finite = values[np.isfinite(values)] + if finite.size == 0: + return 1.0 + spread = float(np.nanmax(finite) - np.nanmin(finite)) + rms = float(np.sqrt(np.mean(finite**2))) + return max(spread, rms, 1e-6) + + +def _coefficients(profile: str) -> dict[str, float]: + normalized = str(profile or "low").strip().lower() + if normalized == "low": + return LOW + if normalized == "high": + return HIGH + raise ValueError(f"Unknown noise profile '{profile}'. Expected 'low' or 'high'.") + + +def _smooth_series(values: np.ndarray) -> np.ndarray: + kernel = np.array([0.15, 0.35, 0.35, 0.15], dtype=float) + return np.convolve(values, kernel, mode="same") + + +def _drift(rng: np.random.Generator, n: int, scale: float) -> np.ndarray: + if n <= 0 or scale <= 0.0: + return np.zeros(n, dtype=float) + raw = rng.normal(0.0, scale, size=n) + return _smooth_series(_smooth_series(raw)) + + +def _event_mask(speed: np.ndarray) -> np.ndarray: + n = speed.size + if n == 0: + return np.zeros(0, dtype=bool) + out = np.zeros(n, dtype=bool) + finite = np.isfinite(speed) + for idx in range(1, n): + if not (finite[idx] and finite[idx - 1]): + continue + if abs(float(speed[idx])) < 1e-6 or abs(float(speed[idx - 1])) < 1e-6: + continue + if np.sign(speed[idx]) != np.sign(speed[idx - 1]): + lo = max(0, idx - 2) + hi = min(n, idx + 3) + out[lo:hi] = True + return out + + +def _force_event_mask(force: np.ndarray) -> np.ndarray: + n = force.size + if n == 0: + return np.zeros(0, dtype=bool) + out = np.zeros(n, dtype=bool) + finite = np.isfinite(force) + if not finite.any(): + return out + peak = float(np.nanmax(np.abs(force[finite]))) + if peak <= 0.0: + return out + impact_indices = np.flatnonzero(finite & (np.abs(force) > 0.01 * peak)) + for idx in impact_indices: + lo = max(0, int(idx) - 1) + hi = min(n, int(idx) + 2) + out[lo:hi] = True + return out + + +def _add_noise_once(df: pd.DataFrame, seed: int = 0, profile: str = "low") -> pd.DataFrame: + coeffs = _coefficients(profile) + out = df.copy() + if ( + _POSITION_COLUMN not in out.columns + and _SPEED_COLUMN not in out.columns + and _FORCE_COLUMN not in out.columns + ): + return out + + speed = _values(out, _SPEED_COLUMN) if _SPEED_COLUMN in out.columns else np.array([], dtype=float) + events = _event_mask(speed) + + if _POSITION_COLUMN in out.columns: + values = _values(out, _POSITION_COLUMN) + scale = _finite_scale(values) + rng = _rng(seed, _POSITION_COLUMN) + noisy = values.copy() + noisy += rng.normal( + 0.0, + coeffs["position_base_sigma_scale"] * scale, + size=values.size, + ) + noisy += _drift(rng, values.size, coeffs["position_drift_sigma_scale"] * scale) + if events.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["position_event_sigma_scale"] * scale, + size=values.size, + ) + * events.astype(float) + ) + quant_step = max( + coeffs["position_quant_step_scale"] * scale, + coeffs["position_quant_step_floor"], + ) + noisy = np.round(noisy / quant_step) * quant_step + out[_POSITION_COLUMN] = noisy + + if _SPEED_COLUMN in out.columns: + values = _values(out, _SPEED_COLUMN) + scale = _finite_scale(values) + rng = _rng(seed, _SPEED_COLUMN) + local_mag = np.abs(values) + ref = float(np.nanmedian(local_mag[np.isfinite(local_mag)])) if np.isfinite(local_mag).any() else 0.0 + noisy = values.copy() + sigma = ( + coeffs["speed_base_sigma_scale"] * scale + + coeffs["speed_hetero_sigma_scale"] * np.maximum(local_mag, ref) + ) + noisy += rng.normal(0.0, sigma, size=values.size) + noisy += _drift(rng, values.size, coeffs["speed_drift_sigma_scale"] * scale) + if events.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["speed_event_sigma_scale"] * scale, + size=values.size, + ) + * events.astype(float) + ) + for idx in np.flatnonzero(events): + if idx + 1 < noisy.size: + noisy[idx + 1] += rng.normal( + 0.0, + coeffs["speed_post_event_sigma_scale"] * scale, + ) + out[_SPEED_COLUMN] = noisy + + if _FORCE_COLUMN in out.columns: + values = _values(out, _FORCE_COLUMN) + scale = _finite_scale(values) + rng = _rng(seed, _FORCE_COLUMN) + force_events = _force_event_mask(values) + noisy = values.copy() + noisy += rng.normal( + 0.0, + coeffs["right_force_base_sigma_scale"] * scale, + size=values.size, + ) + if force_events.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["right_force_event_sigma_scale"] * scale, + size=values.size, + ) + * force_events.astype(float) + ) + out[_FORCE_COLUMN] = noisy + + return out + + +def quantify_noise( + clean: pd.DataFrame, + noisy: pd.DataFrame, + baseline: pd.DataFrame, +) -> dict[str, list[float | str | None]]: + first_diff = first_detectable_time_from_baseline(clean, baseline) + analysis = quantify_analysis( + clean, + noisy, + reference_df=baseline, + first_diff=first_diff, + local_pre_rows=DOCUMENTED_LOCAL_NOISE_ANALYSIS_PRE_ROWS, + local_post_rows=DOCUMENTED_LOCAL_NOISE_ANALYSIS_POST_ROWS, + ) + if first_diff is None or "local" not in analysis: + analysis["local"] = [None] * len(analysis.get("global", [])) + return analysis + + +def add_noise( + clean: pd.DataFrame, + baseline: pd.DataFrame, + seed: int = 0, + noise_level: str = "low", +) -> tuple[pd.DataFrame, dict[str, list[float | str | None]]]: + normalized = str(noise_level or "low").strip().lower() + if normalized not in NOISE_DICT: + raise ValueError(f"Unknown noise level '{noise_level}'. Expected 'low' or 'high'.") + current_seed = int(seed) + for _attempt in range(_MAX_NOISE_RESAMPLE_ATTEMPTS + 1): + noisy_df = _add_noise_once(clean, seed=current_seed, profile=normalized) + noise_analysis = quantify_noise(clean, noisy_df, baseline) + if _analysis_meets_thresholds(noise_analysis, noise_level=normalized): + return noisy_df, noise_analysis + current_seed += 1000 + raise RuntimeError( + f"Could not satisfy minimum SNR thresholds for noise level '{normalized}' " + f"after {_MAX_NOISE_RESAMPLE_ATTEMPTS + 1} attempts." + ) + + +__all__ = ["HIGH", "LOW", "NOISE_DICT", "SNR_THR_DICT", "add_noise", "quantify_noise"] diff --git a/questions/BounceBall/questions.json b/questions/BounceBall/questions.json new file mode 100644 index 0000000000000000000000000000000000000000..7154d755ed52f9617ad06bb9eb7e7d1a3a9c98f2 --- /dev/null +++ b/questions/BounceBall/questions.json @@ -0,0 +1,29274 @@ +{ + "version": 11, + "questions": { + "frost_01234-anchor_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "fern_01234-anchor_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "gentle_01234-anchor_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "frost_01234-cloud_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "fern_01234-cloud_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "gentle_01234-cloud_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "frost_01234-pine_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "fern_01234-pine_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "gentle_01234-pine_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "frost_01234-prairie_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "fern_01234-prairie_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "gentle_01234-prairie_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "frost_01234-spruce_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "fern_01234-spruce_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "gentle_01234-spruce_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "frost_01234-comet_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "fern_01234-comet_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "gentle_01234-comet_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "frost_01234-meadow_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "fern_01234-meadow_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "gentle_01234-meadow_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "frost_01234-river_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "fern_01234-river_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "gentle_01234-river_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "frost_01234-harbor_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "fern_01234-harbor_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "gentle_01234-harbor_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "frost_01234-willow_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "fern_01234-willow_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "gentle_01234-willow_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "frost_01234-flame_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "fern_01234-flame_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "gentle_01234-flame_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "frost_01234-orbit_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "fern_01234-orbit_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "gentle_01234-orbit_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "frost_01234-trail_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "fern_01234-trail_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "gentle_01234-trail_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "frost_01234-island_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "fern_01234-island_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "gentle_01234-island_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "frost_01234-glade_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "fern_01234-glade_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "gentle_01234-glade_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "frost_01234-canyon_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "fern_01234-canyon_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "gentle_01234-canyon_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "frost_01234-ember_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "fern_01234-ember_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "gentle_01234-ember_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "frost_01234-tide_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "fern_01234-tide_0": { + "question_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_1": { + "question_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_2": { + "question_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_3": { + "question_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_4": { + "question_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "gentle_01234-tide_0": { + "question_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_1": { + "question_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_2": { + "question_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_3": { + "question_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_4": { + "question_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "frost_01234-crest_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-star_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-forest_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R.\nThe rail may be tilted by a fixed inclination angle. \nWhen the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail.\nWhen the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-lagoon_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-quartz_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-dawn_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: mass position along the rail\ncol2: mass velocity along the rail\ncol3: force on the right wall\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "coefficient of restitution of the left wall", + "coefficient of restitution of the right wall", + "coefficient of viscous damping", + "inclination angle of the rail", + "weight of the mass", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"coefficient of restitution of the left wall\", \"coefficient of restitution of the right wall\", \"coefficient of viscous damping\", \"inclination angle of the rail\", \"weight of the mass\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-summit_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-aurora_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-valley_0": { + "question_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_1": { + "question_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_2": { + "question_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_3": { + "question_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_4": { + "question_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4", + "label_5" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\", \"label_4\"] denote different parameter changes, while \"label_5\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + } + }, + "label_int_mapping": { + "coefficient of restitution of the left wall": 0, + "coefficient of restitution of the right wall": 1, + "coefficient of viscous damping": 2, + "inclination angle of the rail": 3, + "weight of the mass": 4, + "no parameter change": 5 + }, + "ground_truth_information": { + "pattern_observed": {}, + "interventions": { + "35ce9600f2b649f39794bad12c8451f5": { + "initial_parameters": { + "mass": 0.29498109291776803, + "viscous_damping": 0.001, + "restitution_right": 0.9, + "restitution_left": 0.8, + "initial_position": 3, + "initial_velocity": 10, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "3731111840514a3a84fa9d182c7fb512": { + "initial_parameters": { + "mass": 0.29498109291776803, + "viscous_damping": 0.001, + "restitution_right": 0.9, + "restitution_left": 0.8, + "initial_position": 3, + "initial_velocity": 10, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.3539975269985792, + "first_diff": [ + 17.84000015258789, + 17.780000686645508, + 17.780000686645508 + ] + }, + "d217e2d448755e4bb51130635edf22b9": { + "initial_parameters": { + "mass": 0.29498109291776803, + "viscous_damping": 0.001, + "restitution_right": 0.9, + "restitution_left": 0.8, + "initial_position": 3, + "initial_velocity": 10, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 1.3539422513203463, + "first_diff": [ + 12.600000381469727, + 12.180000305175781, + 17.780000686645508 + ] + }, + "7b377fd33ec55476b4dd35ce0f6adf5b": { + "initial_parameters": { + "mass": 0.29498109291776803, + "viscous_damping": 0.001, + "restitution_right": 0.9, + "restitution_left": 0.8, + "initial_position": 3, + "initial_velocity": 10, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -4, + "first_diff": [ + 12.380000114440918, + 12.0600004196167, + 17.739999771118164 + ] + }, + "55cd6c88ee704897a0318d3680135fb7": { + "initial_parameters": { + "mass": 0.29498109291776803, + "viscous_damping": 0.001, + "restitution_right": 0.9, + "restitution_left": 0.8, + "initial_position": 3, + "initial_velocity": 10, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.3583466866586179, + "first_diff": [ + 13.579999923706055, + 13.579999923706055, + 17.780000686645508 + ] + }, + "81064a33c383447293b0d82c15f891bb": { + "initial_parameters": { + "mass": 0.29498109291776803, + "viscous_damping": 0.001, + "restitution_right": 0.9, + "restitution_left": 0.8, + "initial_position": 3, + "initial_velocity": 10, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 0.37883188366423903, + "first_diff": [ + 12.9399995803833, + 12.680000305175781, + 17.780000686645508 + ] + }, + "e386c1a131f5492f966c7fd71a723ae9": { + "initial_parameters": { + "mass": 8.373359895164388, + "viscous_damping": 0.001, + "restitution_right": 0.7693883204749832, + "restitution_left": 0.7299185268791194, + "initial_position": 3, + "initial_velocity": 3.013609208079666, + "inclination_angle": 0.01 + }, + "intervention_time": 13.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "1aac2ed4a25544c894c596067aeb18ae": { + "initial_parameters": { + "mass": 8.373359895164388, + "viscous_damping": 0.001, + "restitution_right": 0.7693883204749832, + "restitution_left": 0.7299185268791194, + "initial_position": 3, + "initial_velocity": 3.013609208079666, + "inclination_angle": 0.01 + }, + "intervention_time": 13.0, + "changed_parameter": "mass", + "new_value": 48.47654374609071, + "first_diff": [ + 29.899999618530273, + null, + 19.34000015258789 + ] + }, + "5f6f4011eba24b46a09cac472f54fff4": { + "initial_parameters": { + "mass": 8.373359895164388, + "viscous_damping": 0.001, + "restitution_right": 0.7693883204749832, + "restitution_left": 0.7299185268791194, + "initial_position": 3, + "initial_velocity": 3.013609208079666, + "inclination_angle": 0.01 + }, + "intervention_time": 13.0, + "changed_parameter": "restitution_right", + "new_value": 0.3, + "first_diff": [ + 19.520000457763672, + 19.34000015258789, + 19.34000015258789 + ] + }, + "1a6d28e85a7a41a292e5c95ea91ee2cb": { + "initial_parameters": { + "mass": 8.373359895164388, + "viscous_damping": 0.001, + "restitution_right": 0.7693883204749832, + "restitution_left": 0.7299185268791194, + "initial_position": 3, + "initial_velocity": 3.013609208079666, + "inclination_angle": 0.01 + }, + "intervention_time": 20.0, + "changed_parameter": "restitution_left", + "new_value": 0.9999, + "first_diff": [ + 30.860000610351562, + 30.860000610351562, + 42.400001525878906 + ] + }, + "e957c95514144e96beeead74e1ca010a": { + "initial_parameters": { + "mass": 8.373359895164388, + "viscous_damping": 0.001, + "restitution_right": 0.7693883204749832, + "restitution_left": 0.7299185268791194, + "initial_position": 3, + "initial_velocity": 3.013609208079666, + "inclination_angle": 0.01 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.4995105380272218, + "first_diff": [ + 30.860000610351562, + 30.860000610351562, + 46.779998779296875 + ] + }, + "be12a2ede7744ecdae83f9e0d1f2a47f": { + "initial_parameters": { + "mass": 8.373359895164388, + "viscous_damping": 0.001, + "restitution_right": 0.7693883204749832, + "restitution_left": 0.7299185268791194, + "initial_position": 3, + "initial_velocity": 3.013609208079666, + "inclination_angle": 0.01 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.22532772930606587, + "first_diff": [ + 19.5, + 19.34000015258789, + 19.34000015258789 + ] + }, + "664fa7515eea439692b7f460afbf1631": { + "initial_parameters": { + "mass": 8.373359895164388, + "viscous_damping": 0.001, + "restitution_right": 0.7693883204749832, + "restitution_left": 0.7299185268791194, + "initial_position": 3, + "initial_velocity": 3.013609208079666, + "inclination_angle": 0.01 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -0.8693753222674951, + "first_diff": [ + 12.699999809265137, + 12.100000381469727, + 17.81999969482422 + ] + }, + "bbe78dcf65cf4990b574beb1df877cfc": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 15.409, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "98ce3759ae534c9f8ee938e151b01032": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 15.409, + "changed_parameter": "mass", + "new_value": 84.04132573227763, + "first_diff": [ + 30.81999969482422, + 34.31999969482422, + 22.65999984741211 + ] + }, + "12dd1c92dc4649d4808421b7c2dbb124": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 15.409, + "changed_parameter": "restitution_right", + "new_value": 0.3, + "first_diff": [ + 22.84000015258789, + 22.65999984741211, + 47.68000030517578 + ] + }, + "f3359983559e491da7139fb1d378f667": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.5831353947486017, + "first_diff": [ + 15.319999694824219, + 15.319999694824219, + 22.65999984741211 + ] + }, + "2afe39e081a14923a1d46c052479dc3e": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.3607137481284661, + "first_diff": [ + 15.319999694824219, + 15.319999694824219, + 22.65999984741211 + ] + }, + "ebb3d1ed8560449a8e1113f5f0aee8da": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.6422781428784247, + "first_diff": [ + 13.34000015258789, + 12.4399995803833, + 22.65999984741211 + ] + }, + "4e53c84aa5024bf4ae26bb23d7a4d07e": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.2677984719388337, + "first_diff": [ + 22.84000015258789, + 22.65999984741211, + 47.68000030517578 + ] + }, + "22bd886147e847e9a0420134bbfcd2c8": { + "initial_parameters": { + "mass": 28.37626463122252, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.7968370342089809, + "initial_position": 5, + "initial_velocity": -9.129790116443674, + "inclination_angle": 0.003 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -0.86794194573099, + "first_diff": [ + 12.899999618530273, + 12.15999984741211, + 22.65999984741211 + ] + }, + "5990d64f12064184ac3587e5004bac94": { + "initial_parameters": { + "mass": 12, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.9999, + "restitution_left": 0.25, + "initial_position": 9, + "initial_velocity": -4.462262990107625, + "inclination_angle": -0.3522734478477662 + }, + "intervention_time": 18.0, + "changed_parameter": "mass", + "new_value": 62.602441736761534, + "first_diff": [ + 22.200000762939453, + 23.600000381469727, + 42.34000015258789 + ] + }, + "3fd9c7f94e355f65bada44cc585647d6": { + "initial_parameters": { + "mass": 12, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.9999, + "restitution_left": 0.25, + "initial_position": 9, + "initial_velocity": -4.462262990107625, + "inclination_angle": -0.3522734478477662 + }, + "intervention_time": 18.0, + "changed_parameter": "inclination_angle", + "new_value": -2.6255068162366904, + "first_diff": [ + 18.540000915527344, + 18.020000457763672, + 27.739999771118164 + ] + }, + "f9bf7b07efb646ecbc9d989085c45da8": { + "initial_parameters": { + "mass": 12, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.9999, + "restitution_left": 0.25, + "initial_position": 9, + "initial_velocity": -4.462262990107625, + "inclination_angle": -0.3522734478477662 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.5256874925607125, + "first_diff": [ + 23.8799991607666, + 23.8799991607666, + 39.220001220703125 + ] + }, + "4483b5b67b3f4d9abae899f3bea83fa0": { + "initial_parameters": { + "mass": 12, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.9999, + "restitution_left": 0.25, + "initial_position": 9, + "initial_velocity": -4.462262990107625, + "inclination_angle": -0.3522734478477662 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 1.6332726056363913, + "first_diff": [ + 12.899999618530273, + 12.039999961853027, + 12.800000190734863 + ] + }, + "6783ef519fdc4d7bb71f66db00c264e7": { + "initial_parameters": { + "mass": 15.435163116685091, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.83, + "restitution_left": 0.9, + "initial_position": 7, + "initial_velocity": 6.716401440535336, + "inclination_angle": 0.005 + }, + "intervention_time": 10.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "6c4ff7108da34c2d94ba2bb300dd1ee7": { + "initial_parameters": { + "mass": 15.435163116685091, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.83, + "restitution_left": 0.9, + "initial_position": 7, + "initial_velocity": 6.716401440535336, + "inclination_angle": 0.005 + }, + "intervention_time": 10.0, + "changed_parameter": "viscous_damping", + "new_value": 0.001, + "first_diff": [ + 11.0, + 10.279999732971191, + 18.31999969482422 + ] + }, + "245d11d70caf412eb50b80db74cfb5a0": { + "initial_parameters": { + "mass": 15.435163116685091, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.83, + "restitution_left": 0.9, + "initial_position": 7, + "initial_velocity": 6.716401440535336, + "inclination_angle": 0.005 + }, + "intervention_time": 10.0, + "changed_parameter": "restitution_right", + "new_value": 0.3, + "first_diff": [ + 19.940000534057617, + 19.799999237060547, + 19.799999237060547 + ] + }, + "4f26efbd7d274f7fa043588a6d4e7204": { + "initial_parameters": { + "mass": 15.435163116685091, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.83, + "restitution_left": 0.9, + "initial_position": 7, + "initial_velocity": 6.716401440535336, + "inclination_angle": 0.005 + }, + "intervention_time": 10.0, + "changed_parameter": "restitution_left", + "new_value": 0.25, + "first_diff": [ + 12.859999656677246, + 12.859999656677246, + 19.799999237060547 + ] + }, + "e6b5a1b16cc15dbe8d1fe3b51eb03100": { + "initial_parameters": { + "mass": 15.435163116685091, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.83, + "restitution_left": 0.9, + "initial_position": 7, + "initial_velocity": 6.716401440535336, + "inclination_angle": 0.005 + }, + "intervention_time": 10.0, + "changed_parameter": "inclination_angle", + "new_value": -4, + "first_diff": [ + 10.420000076293945, + 10.039999961853027, + 19.799999237060547 + ] + }, + "84bbd6c06ddb4b54a3f12c882c2eccc4": { + "initial_parameters": { + "mass": 22.739065865457274, + "viscous_damping": 0.0141482788417792, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.75, + "initial_position": 5, + "initial_velocity": -10.415361656359881, + "inclination_angle": -0.07675498713894147 + }, + "intervention_time": 14.5, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "7dacaf46a9ec45cf9c7e1aabad31e83f": { + "initial_parameters": { + "mass": 22.739065865457274, + "viscous_damping": 0.0141482788417792, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.75, + "initial_position": 5, + "initial_velocity": -10.415361656359881, + "inclination_angle": -0.07675498713894147 + }, + "intervention_time": 14.5, + "changed_parameter": "mass", + "new_value": 53.0408637333909, + "first_diff": [ + 28.420000076293945, + 35.2599983215332, + 22.479999542236328 + ] + }, + "1b364086c3c740789d832cfebdd6cb1e": { + "initial_parameters": { + "mass": 22.739065865457274, + "viscous_damping": 0.0141482788417792, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.75, + "initial_position": 5, + "initial_velocity": -10.415361656359881, + "inclination_angle": -0.07675498713894147 + }, + "intervention_time": 14.5, + "changed_parameter": "restitution_right", + "new_value": 0.9999, + "first_diff": [ + 22.719999313354492, + 22.5, + 40.31999969482422 + ] + }, + "35428ae77661442ab6f33894965097ef": { + "initial_parameters": { + "mass": 22.739065865457274, + "viscous_damping": 0.0141482788417792, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.75, + "initial_position": 5, + "initial_velocity": -10.415361656359881, + "inclination_angle": -0.07675498713894147 + }, + "intervention_time": 14.5, + "changed_parameter": "restitution_right", + "new_value": 0.3, + "first_diff": [ + 22.700000762939453, + 22.5, + 49.7599983215332 + ] + }, + "4d8fb279215248f8acbfad4f5e395b2e": { + "initial_parameters": { + "mass": 22.739065865457274, + "viscous_damping": 0.0141482788417792, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.75, + "initial_position": 5, + "initial_velocity": -10.415361656359881, + "inclination_angle": -0.07675498713894147 + }, + "intervention_time": 14.5, + "changed_parameter": "restitution_left", + "new_value": 0.10973891356207326, + "first_diff": [ + 14.84000015258789, + 14.84000015258789, + 22.5 + ] + }, + "73e57408da7a53bba190eae95592a58b": { + "initial_parameters": { + "mass": 22.739065865457274, + "viscous_damping": 0.0141482788417792, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.75, + "initial_position": 5, + "initial_velocity": -10.415361656359881, + "inclination_angle": -0.07675498713894147 + }, + "intervention_time": 14.5, + "changed_parameter": "inclination_angle", + "new_value": 1.5, + "first_diff": [ + 14.65999984741211, + 14.579999923706055, + 22.5 + ] + }, + "c7647d63ffa142a181f3bc6d78795879": { + "initial_parameters": { + "mass": 80.93332405513854, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.9999, + "restitution_left": 0.9999, + "initial_position": 4.5, + "initial_velocity": -3.66053324268327, + "inclination_angle": -0.2847467325325903 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "da754a2fc8be40ee99add59263424786": { + "initial_parameters": { + "mass": 80.93332405513854, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.9999, + "restitution_left": 0.9999, + "initial_position": 4.5, + "initial_velocity": -3.66053324268327, + "inclination_angle": -0.2847467325325903 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 12, + "first_diff": [ + 12.640000343322754, + 12.0600004196167, + 13.979999542236328 + ] + }, + "29a0305a7d8947b0983ab26ff8f5b47a": { + "initial_parameters": { + "mass": 80.93332405513854, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.9999, + "restitution_left": 0.9999, + "initial_position": 4.5, + "initial_velocity": -3.66053324268327, + "inclination_angle": -0.2847467325325903 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.3, + "first_diff": [ + 14.039999961853027, + 13.979999542236328, + 13.979999542236328 + ] + }, + "2888456b9ece4be3bd11a8cc7dbf349b": { + "initial_parameters": { + "mass": 80.93332405513854, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.9999, + "restitution_left": 0.9999, + "initial_position": 4.5, + "initial_velocity": -3.66053324268327, + "inclination_angle": -0.2847467325325903 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.25, + "first_diff": [ + 18.459999084472656, + 18.459999084472656, + 23.079999923706055 + ] + }, + "78553d66439e5f678fe0b5574efa7384": { + "initial_parameters": { + "mass": 80.93332405513854, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.9999, + "restitution_left": 0.9999, + "initial_position": 4.5, + "initial_velocity": -3.66053324268327, + "inclination_angle": -0.2847467325325903 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -4, + "first_diff": [ + 12.539999961853027, + 12.039999961853027, + 13.720000267028809 + ] + }, + "4bdd20ee31ce447eb8b038067bdd66d7": { + "initial_parameters": { + "mass": 80.93332405513854, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.9999, + "restitution_left": 0.9999, + "initial_position": 4.5, + "initial_velocity": -3.66053324268327, + "inclination_angle": -0.2847467325325903 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 3.413002101353187, + "first_diff": [ + 13.460000038146973, + 12.279999732971191, + 13.979999542236328 + ] + }, + "d67bbb7d87d745ad8b5f8871e4ae9a10": { + "initial_parameters": { + "mass": 80.93332405513854, + "viscous_damping": 0.5241482788417792, + "restitution_right": 0.9999, + "restitution_left": 0.9999, + "initial_position": 4.5, + "initial_velocity": -3.66053324268327, + "inclination_angle": -0.2847467325325903 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.3554776133905658, + "first_diff": [ + 14.039999961853027, + 13.979999542236328, + 13.979999542236328 + ] + }, + "11e33a2095244285ad1ae8d000034bf2": { + "initial_parameters": { + "mass": 9.165515137260961, + "viscous_damping": 0.5, + "restitution_right": 0.8, + "restitution_left": 0.8968370342089809, + "initial_position": 9, + "initial_velocity": -17.00957899351604, + "inclination_angle": -0.6718716027414882 + }, + "intervention_time": 10.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "07f55a16d69449fcb096b003f6593691": { + "initial_parameters": { + "mass": 9.165515137260961, + "viscous_damping": 0.5, + "restitution_right": 0.8, + "restitution_left": 0.8968370342089809, + "initial_position": 9, + "initial_velocity": -17.00957899351604, + "inclination_angle": -0.6718716027414882 + }, + "intervention_time": 10.0, + "changed_parameter": "mass", + "new_value": 28.607614254023325, + "first_diff": [ + 11.220000267028809, + 10.260000228881836, + 19.459999084472656 + ] + }, + "a04e965c08014d239d8b8a393e1a3cdd": { + "initial_parameters": { + "mass": 9.165515137260961, + "viscous_damping": 0.5, + "restitution_right": 0.8, + "restitution_left": 0.8968370342089809, + "initial_position": 9, + "initial_velocity": -17.00957899351604, + "inclination_angle": -0.6718716027414882 + }, + "intervention_time": 10.0, + "changed_parameter": "viscous_damping", + "new_value": 0.001, + "first_diff": [ + 11.020000457763672, + 10.180000305175781, + 18.760000228881836 + ] + }, + "779261e592034cf882881c2f70a20728": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "ca8e6694b045476e8d251f55f5445ab3": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 58.17152951550182, + "first_diff": [ + 15.140000343322754, + 22.299999237060547, + 22.299999237060547 + ] + }, + "e5d0ca629a2e49bcb76945a581cd1a98": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.9999, + "first_diff": [ + 22.5, + 22.34000015258789, + 22.34000015258789 + ] + }, + "324f1712b21d47888e27b3d794abc71b": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.3, + "first_diff": [ + 22.959999084472656, + 22.34000015258789, + 42.79999923706055 + ] + }, + "d107128850bd502784cdb511d7efa744": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 4, + "first_diff": [ + 12.5, + 12.039999961853027, + 22.34000015258789 + ] + }, + "fab4f5f11b874ef58e909995690d57ec": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.510073138878193, + "first_diff": [ + 15.359999656677246, + 15.359999656677246, + 22.34000015258789 + ] + }, + "da62c8d5d796459081533a11ff95d0c9": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.5119557233367673, + "first_diff": [ + 13.420000076293945, + 12.460000038146973, + 22.34000015258789 + ] + }, + "266ff8b228ea4c2bbc98b40b4b29ae1e": { + "initial_parameters": { + "mass": 23, + "viscous_damping": 0.02, + "restitution_right": 0.44812553614193307, + "restitution_left": 0.99, + "initial_position": 6, + "initial_velocity": 8.556232207255103, + "inclination_angle": 0.953674608912213 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.8038433803627101, + "first_diff": [ + 22.600000381469727, + 22.34000015258789, + 38.380001068115234 + ] + }, + "354dbe3c5b4b4ef29b2e2a1a0ae5c035": { + "initial_parameters": { + "mass": 99.28720221079634, + "viscous_damping": 0.1, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.9999, + "initial_position": 8, + "initial_velocity": -7.957018708726289, + "inclination_angle": -0.03 + }, + "intervention_time": 17.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "8f80071112174821a3595067c2bf1537": { + "initial_parameters": { + "mass": 99.28720221079634, + "viscous_damping": 0.1, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.9999, + "initial_position": 8, + "initial_velocity": -7.957018708726289, + "inclination_angle": -0.03 + }, + "intervention_time": 17.0, + "changed_parameter": "viscous_damping", + "new_value": 10.594738986895932, + "first_diff": [ + 18.020000457763672, + 17.079999923706055, + 29.940000534057617 + ] + }, + "c67fd00576b04ac2bb3baf818c034ae4": { + "initial_parameters": { + "mass": 99.28720221079634, + "viscous_damping": 0.1, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.9999, + "initial_position": 8, + "initial_velocity": -7.957018708726289, + "inclination_angle": -0.03 + }, + "intervention_time": 17.0, + "changed_parameter": "restitution_left", + "new_value": 0.25, + "first_diff": [ + 23.459999084472656, + 23.459999084472656, + 29.940000534057617 + ] + }, + "866e78a3aee946dfaab06893eac3d553": { + "initial_parameters": { + "mass": 99.28720221079634, + "viscous_damping": 0.1, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.9999, + "initial_position": 8, + "initial_velocity": -7.957018708726289, + "inclination_angle": -0.03 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.6948132711679077, + "first_diff": [ + 12.779999732971191, + 12.779999732971191, + 17.040000915527344 + ] + }, + "105dab6037814698b8c5fb1b7cb0d358": { + "initial_parameters": { + "mass": 99.28720221079634, + "viscous_damping": 0.1, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.9999, + "initial_position": 8, + "initial_velocity": -7.957018708726289, + "inclination_angle": -0.03 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 1.645404604070529, + "first_diff": [ + 12.539999961853027, + 12.640000343322754, + 17.040000915527344 + ] + }, + "ed3abb63ca7e49dca5dbc6f41691f421": { + "initial_parameters": { + "mass": 99.28720221079634, + "viscous_damping": 0.1, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.9999, + "initial_position": 8, + "initial_velocity": -7.957018708726289, + "inclination_angle": -0.03 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.25811483485264025, + "first_diff": [ + 17.139999389648438, + 17.040000915527344, + 29.940000534057617 + ] + }, + "7b2b2d6cce1c444d8fc6bc0be9ac42ee": { + "initial_parameters": { + "mass": 99.28720221079634, + "viscous_damping": 0.1, + "restitution_right": 0.6693883204749832, + "restitution_left": 0.9999, + "initial_position": 8, + "initial_velocity": -7.957018708726289, + "inclination_angle": -0.03 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 0.23013911417860558, + "first_diff": [ + 12.5600004196167, + 14.380000114440918, + 17.040000915527344 + ] + }, + "934a4597164b46a7a2728e1ebf583b32": { + "initial_parameters": { + "mass": 38.64311238136688, + "viscous_damping": 1, + "restitution_right": 0.95, + "restitution_left": 0.8, + "initial_position": 6, + "initial_velocity": 4.274507231810087, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "a71f474e41464187af46ccbd7222cae8": { + "initial_parameters": { + "mass": 38.64311238136688, + "viscous_damping": 1, + "restitution_right": 0.95, + "restitution_left": 0.8, + "initial_position": 6, + "initial_velocity": 4.274507231810087, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 71.04877940469977, + "first_diff": [ + 14.680000305175781, + 12.84000015258789, + 12.100000381469727 + ] + }, + "3b00a79d1fff4b02a8566c05ad618150": { + "initial_parameters": { + "mass": 38.64311238136688, + "viscous_damping": 1, + "restitution_right": 0.95, + "restitution_left": 0.8, + "initial_position": 6, + "initial_velocity": 4.274507231810087, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.001, + "first_diff": [ + 13.920000076293945, + 12.380000114440918, + 26.959999084472656 + ] + }, + "5838d4e3197541219e46a399693b4d50": { + "initial_parameters": { + "mass": 38.64311238136688, + "viscous_damping": 1, + "restitution_right": 0.95, + "restitution_left": 0.8, + "initial_position": 6, + "initial_velocity": 4.274507231810087, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.37891066237662796, + "first_diff": [ + 19.360000610351562, + 19.360000610351562, + 30.920000076293945 + ] + }, + "2310e88aa2a541b4a3a573100f28dcad": { + "initial_parameters": { + "mass": 38.64311238136688, + "viscous_damping": 1, + "restitution_right": 0.95, + "restitution_left": 0.8, + "initial_position": 6, + "initial_velocity": 4.274507231810087, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 14.225360588588053, + "first_diff": [ + 13.5600004196167, + 12.220000267028809, + 12.100000381469727 + ] + }, + "74b7a23252a94ac4aa7aa071d7b17aa5": { + "initial_parameters": { + "mass": 38.64311238136688, + "viscous_damping": 1, + "restitution_right": 0.95, + "restitution_left": 0.8, + "initial_position": 6, + "initial_velocity": 4.274507231810087, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.6679477941898871, + "first_diff": [ + 12.300000190734863, + 12.100000381469727, + 30.920000076293945 + ] + }, + "0272b7952029539d9bd40ce24d621991": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.001, + "restitution_right": 0.7, + "restitution_left": 0.9999, + "initial_position": 4.25, + "initial_velocity": -1.8250188146928377, + "inclination_angle": -0.05839057086210227 + }, + "intervention_time": 20.75, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "148cb65464985cf09acc84595e7a7ae3": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.001, + "restitution_right": 0.7, + "restitution_left": 0.9999, + "initial_position": 4.25, + "initial_velocity": -1.8250188146928377, + "inclination_angle": -0.05839057086210227 + }, + "intervention_time": 20.75, + "changed_parameter": "restitution_right", + "new_value": 0.4937070083007543, + "first_diff": [ + 36.20000076293945, + 35.63999938964844, + null + ] + }, + "2f760a132640578b96960b7cee8ae277": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.001, + "restitution_right": 0.7, + "restitution_left": 0.9999, + "initial_position": 4.25, + "initial_velocity": -1.8250188146928377, + "inclination_angle": -0.05839057086210227 + }, + "intervention_time": 20.75, + "changed_parameter": "restitution_left", + "new_value": 0.25, + "first_diff": [ + 22.959999084472656, + 22.959999084472656, + 35.63999938964844 + ] + }, + "8399e56763da5bc59ef9f9c4b652eeee": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.001, + "restitution_right": 0.7, + "restitution_left": 0.9999, + "initial_position": 4.25, + "initial_velocity": -1.8250188146928377, + "inclination_angle": -0.05839057086210227 + }, + "intervention_time": 20.75, + "changed_parameter": "inclination_angle", + "new_value": -4, + "first_diff": [ + 21.0, + 20.780000686645508, + 28.719999313354492 + ] + }, + "cc080755498055de93cebb72440a86c6": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.93, + "restitution_left": 0.91, + "initial_position": 6.25, + "initial_velocity": -8.324265557237352, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.25, + "first_diff": [ + 13.180000305175781, + 13.180000305175781, + 19.299999237060547 + ] + }, + "9584da5231b4512188b5089c9a0dbc05": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.93, + "restitution_left": 0.91, + "initial_position": 6.25, + "initial_velocity": -8.324265557237352, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -4, + "first_diff": [ + 12.279999732971191, + 12.039999961853027, + 18.020000457763672 + ] + }, + "d31bac7398034409ad2a8d05cf5ea3f8": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.93, + "restitution_left": 0.91, + "initial_position": 6.25, + "initial_velocity": -8.324265557237352, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.38978925480156823, + "first_diff": [ + 13.180000305175781, + 13.180000305175781, + 19.299999237060547 + ] + }, + "a3d4d1a02f6f4d74a743b6f0212ec691": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.93, + "restitution_left": 0.91, + "initial_position": 6.25, + "initial_velocity": -8.324265557237352, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.5508410955104519, + "first_diff": [ + 13.180000305175781, + 13.180000305175781, + 19.299999237060547 + ] + }, + "a0399f67cb5f4165a4dc00fa72104bb6": { + "initial_parameters": { + "mass": 0.4641588833612778, + "viscous_damping": 0.022894284851066635, + "restitution_right": 0.93, + "restitution_left": 0.91, + "initial_position": 6.25, + "initial_velocity": -8.324265557237352, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.12239694356771291, + "first_diff": [ + 19.3799991607666, + 19.299999237060547, + 19.299999237060547 + ] + }, + "1032bb73c0b55d1ba0c4685bb20f08f6": { + "initial_parameters": { + "mass": 0.9641588833612778, + "viscous_damping": 0.002894284851066635, + "restitution_right": 0.7481255361419331, + "restitution_left": 0.7968370342089809, + "initial_position": 6.5, + "initial_velocity": 5.042015348383224, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "3a901bfee163599cb81f0751418e5ece": { + "initial_parameters": { + "mass": 0.9641588833612778, + "viscous_damping": 0.002894284851066635, + "restitution_right": 0.7481255361419331, + "restitution_left": 0.7968370342089809, + "initial_position": 6.5, + "initial_velocity": 5.042015348383224, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -4, + "first_diff": [ + 12.5600004196167, + 12.020000457763672, + 19.34000015258789 + ] + }, + "105de8c7a17f4728931344dd36a28313": { + "initial_parameters": { + "mass": 0.9641588833612778, + "viscous_damping": 0.002894284851066635, + "restitution_right": 0.7481255361419331, + "restitution_left": 0.7968370342089809, + "initial_position": 6.5, + "initial_velocity": 5.042015348383224, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.26183449570927336, + "first_diff": [ + 17.739999771118164, + 17.739999771118164, + 26.65999984741211 + ] + }, + "c5b764da8f9442ebbce3b0859ed481a1": { + "initial_parameters": { + "mass": 0.9641588833612778, + "viscous_damping": 0.002894284851066635, + "restitution_right": 0.7481255361419331, + "restitution_left": 0.7968370342089809, + "initial_position": 6.5, + "initial_velocity": 5.042015348383224, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.5023565555121244, + "first_diff": [ + 17.739999771118164, + 17.739999771118164, + 26.65999984741211 + ] + }, + "855c0b4b25c947ddada661a7b2138f42": { + "initial_parameters": { + "mass": 0.9641588833612778, + "viscous_damping": 0.002894284851066635, + "restitution_right": 0.7481255361419331, + "restitution_left": 0.7968370342089809, + "initial_position": 6.5, + "initial_velocity": 5.042015348383224, + "inclination_angle": 0 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 0.7927162548726381, + "first_diff": [ + 13.180000305175781, + 12.15999984741211, + 26.520000457763672 + ] + }, + "bb81df685f9a58e59ecb171a9cceb33e": { + "initial_parameters": { + "mass": 1.35, + "viscous_damping": 0.018, + "restitution_right": 0.7, + "restitution_left": 0.66, + "initial_position": 7.4, + "initial_velocity": -8.9, + "inclination_angle": -0.22 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "b8b64d9c72a35d9486886e9b726261cf": { + "initial_parameters": { + "mass": 1.35, + "viscous_damping": 0.018, + "restitution_right": 0.7, + "restitution_left": 0.66, + "initial_position": 7.4, + "initial_velocity": -8.9, + "inclination_angle": -0.22 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 2.9700000000000006, + "first_diff": [ + 16.68000030517578, + 13.380000114440918, + 13.520000457763672 + ] + }, + "c32fcada64fe5036a1a843f7f423f6b4": { + "initial_parameters": { + "mass": 1.35, + "viscous_damping": 0.018, + "restitution_right": 0.7, + "restitution_left": 0.66, + "initial_position": 7.4, + "initial_velocity": -8.9, + "inclination_angle": -0.22 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.4700912569448683, + "first_diff": [ + 13.779999732971191, + 13.520000457763672, + 40.2599983215332 + ] + }, + "9a03fccac5e1443fbea77e7243b73b84": { + "initial_parameters": { + "mass": 1.35, + "viscous_damping": 0.018, + "restitution_right": 0.7, + "restitution_left": 0.66, + "initial_position": 7.4, + "initial_velocity": -8.9, + "inclination_angle": -0.22 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.3471162860716772, + "first_diff": [ + 24.239999771118164, + 24.239999771118164, + 40.2599983215332 + ] + }, + "25929fbc910e457d9143e2899419263b": { + "initial_parameters": { + "mass": 1.35, + "viscous_damping": 0.018, + "restitution_right": 0.7, + "restitution_left": 0.66, + "initial_position": 7.4, + "initial_velocity": -8.9, + "inclination_angle": -0.22 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.24268559247624535, + "first_diff": [ + 24.239999771118164, + 24.239999771118164, + 40.2599983215332 + ] + }, + "4f60288e7a90582a841638ff2aaca5cc": { + "initial_parameters": { + "mass": 3.1, + "viscous_damping": 0.055, + "restitution_right": 0.82, + "restitution_left": 0.74, + "initial_position": 6.2, + "initial_velocity": -8.2, + "inclination_angle": -0.31 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 6.820000000000001, + "first_diff": [ + 14.420000076293945, + 13.020000457763672, + 26.100000381469727 + ] + }, + "3ff3d291e19c586a958adb1f94864c17": { + "initial_parameters": { + "mass": 3.1, + "viscous_damping": 0.055, + "restitution_right": 0.82, + "restitution_left": 0.74, + "initial_position": 6.2, + "initial_velocity": -8.2, + "inclination_angle": -0.31 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.4135910598853271, + "first_diff": [ + 27.280000686645508, + 27.079999923706055, + 52.040000915527344 + ] + }, + "062a37da2b3c523ea65fe50198329901": { + "initial_parameters": { + "mass": 3.1, + "viscous_damping": 0.055, + "restitution_right": 0.82, + "restitution_left": 0.74, + "initial_position": 6.2, + "initial_velocity": -8.2, + "inclination_angle": -0.31 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 0.99, + "first_diff": [ + 12.960000038146973, + 12.100000381469727, + 27.079999923706055 + ] + }, + "c3b3ddcba6ff5526bd21d1d06ee66506": { + "initial_parameters": { + "mass": 3.1, + "viscous_damping": 0.055, + "restitution_right": 0.82, + "restitution_left": 0.74, + "initial_position": 6.2, + "initial_velocity": -8.2, + "inclination_angle": -0.31 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -1.61, + "first_diff": [ + 12.960000038146973, + 12.100000381469727, + 27.079999923706055 + ] + }, + "1e6654d17d454951848ea2e61337a186": { + "initial_parameters": { + "mass": 3.1, + "viscous_damping": 0.055, + "restitution_right": 0.82, + "restitution_left": 0.74, + "initial_position": 6.2, + "initial_velocity": -8.2, + "inclination_angle": -0.31 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.23517831509701326, + "first_diff": [ + 17.739999771118164, + 17.739999771118164, + 27.079999923706055 + ] + }, + "c131520e5f80470ebb16965a0c6ba8f4": { + "initial_parameters": { + "mass": 3.1, + "viscous_damping": 0.055, + "restitution_right": 0.82, + "restitution_left": 0.74, + "initial_position": 6.2, + "initial_velocity": -8.2, + "inclination_angle": -0.31 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.48641100443514806, + "first_diff": [ + 27.31999969482422, + 27.079999923706055, + 56.099998474121094 + ] + }, + "8549e70b7f975ae68b059707d6673904": { + "initial_parameters": { + "mass": 4.7, + "viscous_damping": 0.095, + "restitution_right": 0.88, + "restitution_left": 0.78, + "initial_position": 5.6, + "initial_velocity": -9.8, + "inclination_angle": 0.34 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "5af136001c8e55ed96639fb169b820df": { + "initial_parameters": { + "mass": 4.7, + "viscous_damping": 0.095, + "restitution_right": 0.88, + "restitution_left": 0.78, + "initial_position": 5.6, + "initial_velocity": -9.8, + "inclination_angle": 0.34 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.28500000000000003, + "first_diff": [ + 12.5, + 12.239999771118164, + 18.020000457763672 + ] + }, + "3f14e6ab62885cecad3876e7b3caf6dd": { + "initial_parameters": { + "mass": 4.7, + "viscous_damping": 0.095, + "restitution_right": 0.88, + "restitution_left": 0.78, + "initial_position": 5.6, + "initial_velocity": -9.8, + "inclination_angle": 0.34 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 1.6400000000000001, + "first_diff": [ + 12.779999732971191, + 12.539999961853027, + 18.020000457763672 + ] + }, + "e4cf1fe8d4165cb1845b624096980de5": { + "initial_parameters": { + "mass": 4.7, + "viscous_damping": 0.095, + "restitution_right": 0.88, + "restitution_left": 0.78, + "initial_position": 5.6, + "initial_velocity": -9.8, + "inclination_angle": 0.34 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -0.96, + "first_diff": [ + 12.779999732971191, + 12.539999961853027, + 17.020000457763672 + ] + }, + "1276239e6f2a43c58fe7ab267bd41d94": { + "initial_parameters": { + "mass": 4.7, + "viscous_damping": 0.095, + "restitution_right": 0.88, + "restitution_left": 0.78, + "initial_position": 5.6, + "initial_velocity": -9.8, + "inclination_angle": 0.34 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_left", + "new_value": 0.4069107080923213, + "first_diff": [ + 12.220000267028809, + 12.220000267028809, + 18.020000457763672 + ] + }, + "9820ec472b7e4371a4d40be054b4c6ae": { + "initial_parameters": { + "mass": 4.7, + "viscous_damping": 0.095, + "restitution_right": 0.88, + "restitution_left": 0.78, + "initial_position": 5.6, + "initial_velocity": -9.8, + "inclination_angle": 0.34 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.20147497167853104, + "first_diff": [ + 12.779999732971191, + 12.4399995803833, + 18.020000457763672 + ] + }, + "2c5b25ed588c52f5a5b4f58fc232e4bb": { + "initial_parameters": { + "mass": 7.2, + "viscous_damping": 0.16, + "restitution_right": 0.54, + "restitution_left": 0.68, + "initial_position": 9.3, + "initial_velocity": -7.7, + "inclination_angle": -0.38 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "95ed9fef81a25479881312e047965573": { + "initial_parameters": { + "mass": 7.2, + "viscous_damping": 0.16, + "restitution_right": 0.54, + "restitution_left": 0.68, + "initial_position": 9.3, + "initial_velocity": -7.7, + "inclination_angle": -0.38 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 15.840000000000002, + "first_diff": [ + 14.420000076293945, + 12.84000015258789, + 20.780000686645508 + ] + }, + "68c17889e0f146f7a0772ecd1c0e63cf": { + "initial_parameters": { + "mass": 7.2, + "viscous_damping": 0.16, + "restitution_right": 0.54, + "restitution_left": 0.68, + "initial_position": 9.3, + "initial_velocity": -7.7, + "inclination_angle": -0.38 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 0.3650309394722907, + "first_diff": [ + 12.579999923706055, + 12.100000381469727, + 21.18000030517578 + ] + }, + "0dda565c44d35ddc944cae93e806df1d": { + "initial_parameters": { + "mass": 10.8, + "viscous_damping": 0.27, + "restitution_right": 0.6, + "restitution_left": 0.72, + "initial_position": 8.6, + "initial_velocity": -8.4, + "inclination_angle": 0.42 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "5ed1c131e780539b84a3e3c3e76d73d4": { + "initial_parameters": { + "mass": 10.8, + "viscous_damping": 0.27, + "restitution_right": 0.6, + "restitution_left": 0.72, + "initial_position": 8.6, + "initial_velocity": -8.4, + "inclination_angle": 0.42 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 23.760000000000005, + "first_diff": [ + 15.380000114440918, + 12.720000267028809, + 16.440000534057617 + ] + }, + "ba1a51d50f0b5031b0f10f2b48f8f5b5": { + "initial_parameters": { + "mass": 10.8, + "viscous_damping": 0.27, + "restitution_right": 0.6, + "restitution_left": 0.72, + "initial_position": 8.6, + "initial_velocity": -8.4, + "inclination_angle": 0.42 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.81, + "first_diff": [ + 13.539999961853027, + 12.199999809265137, + 16.6200008392334 + ] + }, + "45aa02ea71c0510aae5f198aef30fef5": { + "initial_parameters": { + "mass": 10.8, + "viscous_damping": 0.27, + "restitution_right": 0.6, + "restitution_left": 0.72, + "initial_position": 8.6, + "initial_velocity": -8.4, + "inclination_angle": 0.42 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.09000000000000001, + "first_diff": [ + 14.960000038146973, + 12.579999923706055, + 16.399999618530273 + ] + }, + "42a04eb48f2359898f15b1581dab6727": { + "initial_parameters": { + "mass": 10.8, + "viscous_damping": 0.27, + "restitution_right": 0.6, + "restitution_left": 0.72, + "initial_position": 8.6, + "initial_velocity": -8.4, + "inclination_angle": 0.42 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 1.72, + "first_diff": [ + 12.899999618530273, + 12.079999923706055, + 16.6200008392334 + ] + }, + "6ca1f93c5f9d59669a10db584518a5da": { + "initial_parameters": { + "mass": 10.8, + "viscous_damping": 0.27, + "restitution_right": 0.6, + "restitution_left": 0.72, + "initial_position": 8.6, + "initial_velocity": -8.4, + "inclination_angle": 0.42 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -0.8800000000000001, + "first_diff": [ + 12.899999618530273, + 12.079999923706055, + 15.600000381469727 + ] + }, + "e9cc36d951655bdfbcf90d0bb2322a0e": { + "initial_parameters": { + "mass": 16.5, + "viscous_damping": 0.46, + "restitution_right": 0.66, + "restitution_left": 0.76, + "initial_position": 7.9, + "initial_velocity": -9.1, + "inclination_angle": -0.47 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "90872800e402540eb77c39d26a123b23": { + "initial_parameters": { + "mass": 16.5, + "viscous_damping": 0.46, + "restitution_right": 0.66, + "restitution_left": 0.76, + "initial_position": 7.9, + "initial_velocity": -9.1, + "inclination_angle": -0.47 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 36.300000000000004, + "first_diff": [ + 14.979999542236328, + 12.640000343322754, + 12.460000038146973 + ] + }, + "9698d8207595571e9187650324239bb0": { + "initial_parameters": { + "mass": 16.5, + "viscous_damping": 0.46, + "restitution_right": 0.66, + "restitution_left": 0.76, + "initial_position": 7.9, + "initial_velocity": -9.1, + "inclination_angle": -0.47 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 7.499999999999999, + "first_diff": [ + 13.460000038146973, + 12.300000190734863, + 12.460000038146973 + ] + }, + "be961034c82352e09e4c5952be16637a": { + "initial_parameters": { + "mass": 16.5, + "viscous_damping": 0.46, + "restitution_right": 0.66, + "restitution_left": 0.76, + "initial_position": 7.9, + "initial_velocity": -9.1, + "inclination_angle": -0.47 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 1.3800000000000001, + "first_diff": [ + 13.260000228881836, + 12.180000305175781, + 12.460000038146973 + ] + }, + "6be5dc20d4365c68ae8ef3bae8a17615": { + "initial_parameters": { + "mass": 16.5, + "viscous_damping": 0.46, + "restitution_right": 0.66, + "restitution_left": 0.76, + "initial_position": 7.9, + "initial_velocity": -9.1, + "inclination_angle": -0.47 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.15333333333333335, + "first_diff": [ + 14.739999771118164, + 12.539999961853027, + 38.619998931884766 + ] + }, + "96f1621029835fd49654f3249386c245": { + "initial_parameters": { + "mass": 16.5, + "viscous_damping": 0.46, + "restitution_right": 0.66, + "restitution_left": 0.76, + "initial_position": 7.9, + "initial_velocity": -9.1, + "inclination_angle": -0.47 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.48000000000000004, + "first_diff": [ + 12.739999771118164, + 12.460000038146973, + 41.20000076293945 + ] + }, + "e6614388019a4d22bdd9140403a47547": { + "initial_parameters": { + "mass": 16.5, + "viscous_damping": 0.46, + "restitution_right": 0.66, + "restitution_left": 0.76, + "initial_position": 7.9, + "initial_velocity": -9.1, + "inclination_angle": -0.47 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 5.237449130812625, + "first_diff": [ + 13.220000267028809, + 12.15999984741211, + 12.460000038146973 + ] + }, + "7379873992335258a1c5571e0f7902db": { + "initial_parameters": { + "mass": 25, + "viscous_damping": 0.78, + "restitution_right": 0.72, + "restitution_left": 0.64, + "initial_position": 7.1, + "initial_velocity": -8.8, + "inclination_angle": 0.51 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "e755a27e73ee5af2bf1337a1a71d0ea3": { + "initial_parameters": { + "mass": 25, + "viscous_damping": 0.78, + "restitution_right": 0.72, + "restitution_left": 0.64, + "initial_position": 7.1, + "initial_velocity": -8.8, + "inclination_angle": 0.51 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 55.00000000000001, + "first_diff": [ + 15.100000381469727, + 12.579999923706055, + 17.860000610351562 + ] + }, + "fc653cb5b3c05a259ddecd3547667a15": { + "initial_parameters": { + "mass": 25, + "viscous_damping": 0.78, + "restitution_right": 0.72, + "restitution_left": 0.64, + "initial_position": 7.1, + "initial_velocity": -8.8, + "inclination_angle": 0.51 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.26, + "first_diff": [ + 14.720000267028809, + 12.460000038146973, + 17.780000686645508 + ] + }, + "cc34c415d1465097ae993175b76900b8": { + "initial_parameters": { + "mass": 25, + "viscous_damping": 0.78, + "restitution_right": 0.72, + "restitution_left": 0.64, + "initial_position": 7.1, + "initial_velocity": -8.8, + "inclination_angle": 0.51 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -0.79, + "first_diff": [ + 12.880000114440918, + 12.0600004196167, + 16.239999771118164 + ] + }, + "b248181b23624fdabde51fb023f87d70": { + "initial_parameters": { + "mass": 25, + "viscous_damping": 0.78, + "restitution_right": 0.72, + "restitution_left": 0.64, + "initial_position": 7.1, + "initial_velocity": -8.8, + "inclination_angle": 0.51 + }, + "intervention_time": 12.0, + "changed_parameter": "restitution_right", + "new_value": 0.17633265595674133, + "first_diff": [ + 18.68000030517578, + 18.34000015258789, + 18.34000015258789 + ] + }, + "3828621236495616b794ce84cf8c84ea": { + "initial_parameters": { + "mass": 38, + "viscous_damping": 1.32, + "restitution_right": 0.78, + "restitution_left": 0.6, + "initial_position": 6.4, + "initial_velocity": -9.4, + "inclination_angle": -0.56 + }, + "intervention_time": 12.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "acd65eb4f78a5539a10bb7f268c00b8b": { + "initial_parameters": { + "mass": 38, + "viscous_damping": 1.32, + "restitution_right": 0.78, + "restitution_left": 0.6, + "initial_position": 6.4, + "initial_velocity": -9.4, + "inclination_angle": -0.56 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 83.60000000000001, + "first_diff": [ + 14.720000267028809, + 12.520000457763672, + 15.020000457763672 + ] + }, + "2736666ef08051b18037717bad48edcc": { + "initial_parameters": { + "mass": 38, + "viscous_damping": 1.32, + "restitution_right": 0.78, + "restitution_left": 0.6, + "initial_position": 6.4, + "initial_velocity": -9.4, + "inclination_angle": -0.56 + }, + "intervention_time": 12.0, + "changed_parameter": "mass", + "new_value": 17.27272727272727, + "first_diff": [ + 13.699999809265137, + 12.239999771118164, + 15.100000381469727 + ] + }, + "6407da2f934a51c5951cb7161ace6674": { + "initial_parameters": { + "mass": 38, + "viscous_damping": 1.32, + "restitution_right": 0.78, + "restitution_left": 0.6, + "initial_position": 6.4, + "initial_velocity": -9.4, + "inclination_angle": -0.56 + }, + "intervention_time": 12.0, + "changed_parameter": "viscous_damping", + "new_value": 0.44, + "first_diff": [ + 14.399999618530273, + 12.420000076293945, + 15.0 + ] + }, + "0310deb38f465db18ad47c098129e23d": { + "initial_parameters": { + "mass": 38, + "viscous_damping": 1.32, + "restitution_right": 0.78, + "restitution_left": 0.6, + "initial_position": 6.4, + "initial_velocity": -9.4, + "inclination_angle": -0.56 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": 0.74, + "first_diff": [ + 12.960000038146973, + 12.079999923706055, + 15.100000381469727 + ] + }, + "461103bd3ac255b2a1e6bd5a4122cbc4": { + "initial_parameters": { + "mass": 38, + "viscous_damping": 1.32, + "restitution_right": 0.78, + "restitution_left": 0.6, + "initial_position": 6.4, + "initial_velocity": -9.4, + "inclination_angle": -0.56 + }, + "intervention_time": 12.0, + "changed_parameter": "inclination_angle", + "new_value": -1.86, + "first_diff": [ + 12.960000038146973, + 12.079999923706055, + 14.720000267028809 + ] + } + } + }, + "environment_name": "DampedMassBetweenWalls" +} \ No newline at end of file diff --git a/questions/BounceBall/sample_manifest.json b/questions/BounceBall/sample_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..0b1da4d253226970e52294afeeb6e3f826b106ef --- /dev/null +++ b/questions/BounceBall/sample_manifest.json @@ -0,0 +1,2918 @@ +{ + "frost_01234": [ + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "c7647d63ffa142a181f3bc6d78795879", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "8549e70b7f975ae68b059707d6673904", + "f12e13839aa35418b517ce7c64e0da71", + "934a4597164b46a7a2728e1ebf583b32", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "3828621236495616b794ce84cf8c84ea", + "779261e592034cf882881c2f70a20728", + "bbe78dcf65cf4990b574beb1df877cfc" + ], + "test_samples": [ + "2888456b9ece4be3bd11a8cc7dbf349b", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3731111840514a3a84fa9d182c7fb512", + "3f14e6ab62885cecad3876e7b3caf6dd", + "4f60288e7a90582a841638ff2aaca5cc", + "5838d4e3197541219e46a399693b4d50", + "68c17889e0f146f7a0772ecd1c0e63cf", + "acd65eb4f78a5539a10bb7f268c00b8b", + "e5d0ca629a2e49bcb76945a581cd1a98", + "ebb3d1ed8560449a8e1113f5f0aee8da" + ], + "test_samples_hashes": [ + "2b7e346a96d4715c8f02067c623fdc62", + "243f64406b1770defea4dfd9855d4090", + "d70d86fcdb1ed345b6cf75172957e4f7", + "f6112bf6818c85259ec05f4aef8013d6", + "843f2e5668569c0c4c2a29931dc3b2a9", + "83cc529948817b70b62e5338cac2f443", + "3681ea0983bf36d9744aa3516845b1b1", + "aa813762880b0c48ade181b6ffc371fb", + "7affebd7a60baafa0c4cd5e84b9ef2ef", + "76f91b400bf19af2e7734c0a010fe808" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "3ebf9b15a2d2c7a674f4931e93289d86e92772e40768276bcc90ada7bf7d765f", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "35ce9600f2b649f39794bad12c8451f5", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "55cd6c88ee704897a0318d3680135fb7", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "779261e592034cf882881c2f70a20728", + "934a4597164b46a7a2728e1ebf583b32", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "e386c1a131f5492f966c7fd71a723ae9", + "35ce9600f2b649f39794bad12c8451f5", + "8549e70b7f975ae68b059707d6673904", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "6783ef519fdc4d7bb71f66db00c264e7", + "7379873992335258a1c5571e0f7902db" + ], + "test_samples": [ + "324f1712b21d47888e27b3d794abc71b", + "3b00a79d1fff4b02a8566c05ad618150", + "4d8fb279215248f8acbfad4f5e395b2e", + "5f6f4011eba24b46a09cac472f54fff4", + "81064a33c383447293b0d82c15f891bb", + "8549e70b7f975ae68b059707d6673904", + "d31bac7398034409ad2a8d05cf5ea3f8", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3" + ], + "test_samples_hashes": [ + "635fe28f1f1a58be42225a9b61a1d56c", + "a291f1aabdec43557934be734a6a69ce", + "11c88f49a40db5a915e37952eb0d7053", + "378ac60f3f79eac1e573015cfe1d5d38", + "a25869b63aeea4d24d4b36d2ffb62d10", + "5f3821630bc45ce5aa5333059a160167", + "1e71612c7615089a03357efd6bf8ec9f", + "0f603778fccc9f44ee65fb6bf8140069", + "6028a70fcdc33ddaf14ddd1e660f34d3", + "21262e359a3b624c46be559d7db5aae5" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "251c9e9d6c7f1d2432a7859586530735e5b616805aebc545831ea47e6a16e733", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.3, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "f12e13839aa35418b517ce7c64e0da71", + "35ce9600f2b649f39794bad12c8451f5", + "934a4597164b46a7a2728e1ebf583b32", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "bbe78dcf65cf4990b574beb1df877cfc", + "bb81df685f9a58e59ecb171a9cceb33e", + "11e33a2095244285ad1ae8d000034bf2", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "779261e592034cf882881c2f70a20728", + "8549e70b7f975ae68b059707d6673904" + ], + "test_samples": [ + "062a37da2b3c523ea65fe50198329901", + "35ce9600f2b649f39794bad12c8451f5", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "96f1621029835fd49654f3249386c245", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a04e965c08014d239d8b8a393e1a3cdd", + "c5b764da8f9442ebbce3b0859ed481a1", + "ca8e6694b045476e8d251f55f5445ab3", + "e4cf1fe8d4165cb1845b624096980de5" + ], + "test_samples_hashes": [ + "e25f2297ed37ce231a36f6642322d064", + "b054165364e7ab8827bbcaf351c56534", + "795785f4ba2c82450626bf9385860036", + "5d776197ac4eab709532b359e5c94428", + "e9eca74cd5d96b1881f09c2085afb794", + "63687f5fc9371aefd313ed10992bd199", + "3572dd7ed11afe351f00c15cb16e62f1", + "ed4fb51fbbe9085bcc26ab5ec9f832a4", + "df4762fc4837017b44d02f3d9712953f", + "149c92837125f8dc1559cabfcc6f11d2" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "de527794b8e5af99d78d21c133542e454991eaeb18bf7fb8df68e06d8f1bb33e", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "9820ec472b7e4371a4d40be054b4c6ae", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3828621236495616b794ce84cf8c84ea", + "0272b7952029539d9bd40ce24d621991", + "0dda565c44d35ddc944cae93e806df1d", + "bbe78dcf65cf4990b574beb1df877cfc", + "35ce9600f2b649f39794bad12c8451f5", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "934a4597164b46a7a2728e1ebf583b32", + "bb81df685f9a58e59ecb171a9cceb33e", + "c7647d63ffa142a181f3bc6d78795879" + ], + "test_samples": [ + "105dab6037814698b8c5fb1b7cb0d358", + "2736666ef08051b18037717bad48edcc", + "2f760a132640578b96960b7cee8ae277", + "42a04eb48f2359898f15b1581dab6727", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "bb81df685f9a58e59ecb171a9cceb33e", + "d67bbb7d87d745ad8b5f8871e4ae9a10" + ], + "test_samples_hashes": [ + "b8b3a9bced67d7217431b9939f2a9078", + "9c9d519dabd56d172e09e05754193d47", + "df46b10959e0653936736c9ad43bbfe1", + "09d201c2304591700da298bc6e543570", + "035d38b8be32e0e01aab8449ac8c23d3", + "f9455494e70645b8c2ea7aba858e9449", + "a362aefef1a80b9fd63926fbb535ce10", + "e5c8469aedae7cbf8b54337f754e5ad4", + "402af21e1c849e04ac412fe2d544fbf5", + "75de90efb4b12116d19217bae9278b00" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "a14fd4de2d8744107c4bc136b418bff031d3821dfe49941f41ff10b2307649ca", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.4, + "euclidean_centroid": 0.2, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "1032bb73c0b55d1ba0c4685bb20f08f6", + "8549e70b7f975ae68b059707d6673904", + "e386c1a131f5492f966c7fd71a723ae9", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "6783ef519fdc4d7bb71f66db00c264e7", + "a17a6c0b50f444b380cd3867d55fee17", + "c7647d63ffa142a181f3bc6d78795879", + "0dda565c44d35ddc944cae93e806df1d", + "779261e592034cf882881c2f70a20728", + "2c5b25ed588c52f5a5b4f58fc232e4bb" + ], + "test_samples": [ + "105de8c7a17f4728931344dd36a28313", + "1276239e6f2a43c58fe7ab267bd41d94", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "245d11d70caf412eb50b80db74cfb5a0", + "4483b5b67b3f4d9abae899f3bea83fa0", + "4bdd20ee31ce447eb8b038067bdd66d7", + "6ca1f93c5f9d59669a10db584518a5da", + "779261e592034cf882881c2f70a20728", + "95ed9fef81a25479881312e047965573" + ], + "test_samples_hashes": [ + "1e55b3e782900756be811ce23aa9c293", + "7a563abc0a8b48d95b01f5708a0c44cd", + "841e717955365a497a1f405c126bd395", + "3fca2198a402cddcc08c6f711ddd1ba5", + "f68a5f99d2430789417bfec590c86048", + "5a4e3bfbfe29a777450556dee062d92f", + "c61b35e73ff1027e36949073efc0f36f", + "5598a5fe92b467882d3a4e316c2ac107", + "f85f5510d85c3d8597a4b8a3fc177afa", + "50f704f5c9b607dded1292f4f7629096" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "2e36527196f792115cc69f11e7b680047b165f4dafc687ebcc22612a27a9ff9b", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "11e33a2095244285ad1ae8d000034bf2", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + } + } + ], + "fern_01234": [ + { + "train_samples_baselines": [ + "11e33a2095244285ad1ae8d000034bf2", + "0272b7952029539d9bd40ce24d621991", + "e386c1a131f5492f966c7fd71a723ae9", + "0dda565c44d35ddc944cae93e806df1d", + "bb81df685f9a58e59ecb171a9cceb33e", + "1032bb73c0b55d1ba0c4685bb20f08f6" + ], + "train_samples": [ + "07f55a16d69449fcb096b003f6593691", + "148cb65464985cf09acc84595e7a7ae3", + "664fa7515eea439692b7f460afbf1631", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "c5b764da8f9442ebbce3b0859ed481a1" + ], + "train_samples_hashes": [ + "82d6de46d567d7269e33254ae97e5438", + "a97fdbdc4907bef0fa4d576e91eafdc0", + "26ae001a83106a1ff76cc174e0797c70", + "8eec429a21e528fd576f160066e454dc", + "402af21e1c849e04ac412fe2d544fbf5", + "ed4fb51fbbe9085bcc26ab5ec9f832a4" + ], + "test_samples_baselines": [ + "c7647d63ffa142a181f3bc6d78795879", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "8549e70b7f975ae68b059707d6673904", + "f12e13839aa35418b517ce7c64e0da71", + "934a4597164b46a7a2728e1ebf583b32", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "3828621236495616b794ce84cf8c84ea", + "779261e592034cf882881c2f70a20728", + "bbe78dcf65cf4990b574beb1df877cfc" + ], + "test_samples": [ + "2888456b9ece4be3bd11a8cc7dbf349b", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3731111840514a3a84fa9d182c7fb512", + "3f14e6ab62885cecad3876e7b3caf6dd", + "4f60288e7a90582a841638ff2aaca5cc", + "5838d4e3197541219e46a399693b4d50", + "68c17889e0f146f7a0772ecd1c0e63cf", + "acd65eb4f78a5539a10bb7f268c00b8b", + "e5d0ca629a2e49bcb76945a581cd1a98", + "ebb3d1ed8560449a8e1113f5f0aee8da" + ], + "test_samples_hashes": [ + "2b7e346a96d4715c8f02067c623fdc62", + "243f64406b1770defea4dfd9855d4090", + "d70d86fcdb1ed345b6cf75172957e4f7", + "f6112bf6818c85259ec05f4aef8013d6", + "843f2e5668569c0c4c2a29931dc3b2a9", + "83cc529948817b70b62e5338cac2f443", + "3681ea0983bf36d9744aa3516845b1b1", + "aa813762880b0c48ade181b6ffc371fb", + "7affebd7a60baafa0c4cd5e84b9ef2ef", + "76f91b400bf19af2e7734c0a010fe808" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "1de0f05d901cd2262af371976539b3b986205d71ef6add839874421f294348a4", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "35ce9600f2b649f39794bad12c8451f5", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "55cd6c88ee704897a0318d3680135fb7", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "6783ef519fdc4d7bb71f66db00c264e7", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "f12e13839aa35418b517ce7c64e0da71", + "0dda565c44d35ddc944cae93e806df1d", + "f12e13839aa35418b517ce7c64e0da71", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "bbe78dcf65cf4990b574beb1df877cfc" + ], + "train_samples": [ + "3ff3d291e19c586a958adb1f94864c17", + "45aa02ea71c0510aae5f198aef30fef5", + "4f60288e7a90582a841638ff2aaca5cc", + "68c17889e0f146f7a0772ecd1c0e63cf", + "866e78a3aee946dfaab06893eac3d553", + "bbe78dcf65cf4990b574beb1df877cfc" + ], + "train_samples_hashes": [ + "c8e0fe58f23de5f3f75383e8f0ea9eee", + "680349aa8ffea35b1f3bc0174b398a0d", + "843f2e5668569c0c4c2a29931dc3b2a9", + "3681ea0983bf36d9744aa3516845b1b1", + "f658b04327b8b0123de2350dc1f9fe23", + "09897ffa321559f9ad73838b55cbba8d" + ], + "test_samples_baselines": [ + "779261e592034cf882881c2f70a20728", + "934a4597164b46a7a2728e1ebf583b32", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "e386c1a131f5492f966c7fd71a723ae9", + "35ce9600f2b649f39794bad12c8451f5", + "8549e70b7f975ae68b059707d6673904", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "6783ef519fdc4d7bb71f66db00c264e7", + "7379873992335258a1c5571e0f7902db" + ], + "test_samples": [ + "324f1712b21d47888e27b3d794abc71b", + "3b00a79d1fff4b02a8566c05ad618150", + "4d8fb279215248f8acbfad4f5e395b2e", + "5f6f4011eba24b46a09cac472f54fff4", + "81064a33c383447293b0d82c15f891bb", + "8549e70b7f975ae68b059707d6673904", + "d31bac7398034409ad2a8d05cf5ea3f8", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3" + ], + "test_samples_hashes": [ + "635fe28f1f1a58be42225a9b61a1d56c", + "a291f1aabdec43557934be734a6a69ce", + "11c88f49a40db5a915e37952eb0d7053", + "378ac60f3f79eac1e573015cfe1d5d38", + "a25869b63aeea4d24d4b36d2ffb62d10", + "5f3821630bc45ce5aa5333059a160167", + "1e71612c7615089a03357efd6bf8ec9f", + "0f603778fccc9f44ee65fb6bf8140069", + "6028a70fcdc33ddaf14ddd1e660f34d3", + "21262e359a3b624c46be559d7db5aae5" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "731e0ac46bb9427cd3802e15884346ae531c97cf8ee63a133c6f8c9603f7b722", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "855c0b4b25c947ddada661a7b2138f42", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.3, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "e386c1a131f5492f966c7fd71a723ae9", + "e386c1a131f5492f966c7fd71a723ae9", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "c7647d63ffa142a181f3bc6d78795879" + ], + "train_samples": [ + "1aac2ed4a25544c894c596067aeb18ae", + "5f6f4011eba24b46a09cac472f54fff4", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "8f80071112174821a3595067c2bf1537", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879" + ], + "train_samples_hashes": [ + "841e717955365a497a1f405c126bd395", + "378ac60f3f79eac1e573015cfe1d5d38", + "e76753a3799574c39b0e86afafa1b2ed", + "0b8ddbea3b36ee43473118eb56d7ae9a", + "586e10e4d484c858801f077398ee7419", + "2d1bf83f69cc4a383d75275447f73aa8" + ], + "test_samples_baselines": [ + "f12e13839aa35418b517ce7c64e0da71", + "35ce9600f2b649f39794bad12c8451f5", + "934a4597164b46a7a2728e1ebf583b32", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "bbe78dcf65cf4990b574beb1df877cfc", + "bb81df685f9a58e59ecb171a9cceb33e", + "11e33a2095244285ad1ae8d000034bf2", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "779261e592034cf882881c2f70a20728", + "8549e70b7f975ae68b059707d6673904" + ], + "test_samples": [ + "062a37da2b3c523ea65fe50198329901", + "35ce9600f2b649f39794bad12c8451f5", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "96f1621029835fd49654f3249386c245", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a04e965c08014d239d8b8a393e1a3cdd", + "c5b764da8f9442ebbce3b0859ed481a1", + "ca8e6694b045476e8d251f55f5445ab3", + "e4cf1fe8d4165cb1845b624096980de5" + ], + "test_samples_hashes": [ + "e25f2297ed37ce231a36f6642322d064", + "b054165364e7ab8827bbcaf351c56534", + "795785f4ba2c82450626bf9385860036", + "5d776197ac4eab709532b359e5c94428", + "e9eca74cd5d96b1881f09c2085afb794", + "63687f5fc9371aefd313ed10992bd199", + "3572dd7ed11afe351f00c15cb16e62f1", + "ed4fb51fbbe9085bcc26ab5ec9f832a4", + "df4762fc4837017b44d02f3d9712953f", + "149c92837125f8dc1559cabfcc6f11d2" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "fdad063ad4044f77ebabb52c3c2048f845087c68c0e47f3694207b5b2d2f1a19", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "9820ec472b7e4371a4d40be054b4c6ae", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "6783ef519fdc4d7bb71f66db00c264e7", + "7379873992335258a1c5571e0f7902db", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "779261e592034cf882881c2f70a20728", + "779261e592034cf882881c2f70a20728", + "a17a6c0b50f444b380cd3867d55fee17" + ], + "train_samples": [ + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "7379873992335258a1c5571e0f7902db", + "9698d8207595571e9187650324239bb0", + "d107128850bd502784cdb511d7efa744", + "e5d0ca629a2e49bcb76945a581cd1a98", + "f9bf7b07efb646ecbc9d989085c45da8" + ], + "train_samples_hashes": [ + "6a0550716d5eefc02e5a3d0ba2a08589", + "1f1a2fb77d47733588a417e23660ac4f", + "c554c189cfd1f44941d4ad41c2d8e289", + "ed5d8307e5f7968b39bcfd819b4d6d20", + "7affebd7a60baafa0c4cd5e84b9ef2ef", + "df68d53d89636da356cfa7e64eb41e90" + ], + "test_samples_baselines": [ + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3828621236495616b794ce84cf8c84ea", + "0272b7952029539d9bd40ce24d621991", + "0dda565c44d35ddc944cae93e806df1d", + "bbe78dcf65cf4990b574beb1df877cfc", + "35ce9600f2b649f39794bad12c8451f5", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "934a4597164b46a7a2728e1ebf583b32", + "bb81df685f9a58e59ecb171a9cceb33e", + "c7647d63ffa142a181f3bc6d78795879" + ], + "test_samples": [ + "105dab6037814698b8c5fb1b7cb0d358", + "2736666ef08051b18037717bad48edcc", + "2f760a132640578b96960b7cee8ae277", + "42a04eb48f2359898f15b1581dab6727", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "bb81df685f9a58e59ecb171a9cceb33e", + "d67bbb7d87d745ad8b5f8871e4ae9a10" + ], + "test_samples_hashes": [ + "b8b3a9bced67d7217431b9939f2a9078", + "9c9d519dabd56d172e09e05754193d47", + "df46b10959e0653936736c9ad43bbfe1", + "09d201c2304591700da298bc6e543570", + "035d38b8be32e0e01aab8449ac8c23d3", + "f9455494e70645b8c2ea7aba858e9449", + "a362aefef1a80b9fd63926fbb535ce10", + "e5c8469aedae7cbf8b54337f754e5ad4", + "402af21e1c849e04ac412fe2d544fbf5", + "75de90efb4b12116d19217bae9278b00" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "161a98da89089f46594fd44e3a9f87138302f5d1520905d4a3b52aa9e54c01f1", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6ca1f93c5f9d59669a10db584518a5da", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.4, + "euclidean_centroid": 0.2, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [ + "bbe78dcf65cf4990b574beb1df877cfc", + "3828621236495616b794ce84cf8c84ea", + "35ce9600f2b649f39794bad12c8451f5", + "11e33a2095244285ad1ae8d000034bf2", + "bb81df685f9a58e59ecb171a9cceb33e", + "7379873992335258a1c5571e0f7902db" + ], + "train_samples": [ + "22bd886147e847e9a0420134bbfcd2c8", + "3828621236495616b794ce84cf8c84ea", + "55cd6c88ee704897a0318d3680135fb7", + "a04e965c08014d239d8b8a393e1a3cdd", + "c32fcada64fe5036a1a843f7f423f6b4", + "e755a27e73ee5af2bf1337a1a71d0ea3" + ], + "train_samples_hashes": [ + "1eb602b03135fca1ba4be3c40f5ed5bf", + "646d80e73dfb4473643ec711415ec877", + "8c8bedd0a9bb68365bf480743ddb2721", + "3572dd7ed11afe351f00c15cb16e62f1", + "8e4b8dd31d946bf8483cb98cc6ca28f4", + "21262e359a3b624c46be559d7db5aae5" + ], + "test_samples_baselines": [ + "1032bb73c0b55d1ba0c4685bb20f08f6", + "8549e70b7f975ae68b059707d6673904", + "e386c1a131f5492f966c7fd71a723ae9", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "6783ef519fdc4d7bb71f66db00c264e7", + "a17a6c0b50f444b380cd3867d55fee17", + "c7647d63ffa142a181f3bc6d78795879", + "0dda565c44d35ddc944cae93e806df1d", + "779261e592034cf882881c2f70a20728", + "2c5b25ed588c52f5a5b4f58fc232e4bb" + ], + "test_samples": [ + "105de8c7a17f4728931344dd36a28313", + "1276239e6f2a43c58fe7ab267bd41d94", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "245d11d70caf412eb50b80db74cfb5a0", + "4483b5b67b3f4d9abae899f3bea83fa0", + "4bdd20ee31ce447eb8b038067bdd66d7", + "6ca1f93c5f9d59669a10db584518a5da", + "779261e592034cf882881c2f70a20728", + "95ed9fef81a25479881312e047965573" + ], + "test_samples_hashes": [ + "1e55b3e782900756be811ce23aa9c293", + "7a563abc0a8b48d95b01f5708a0c44cd", + "841e717955365a497a1f405c126bd395", + "3fca2198a402cddcc08c6f711ddd1ba5", + "f68a5f99d2430789417bfec590c86048", + "5a4e3bfbfe29a777450556dee062d92f", + "c61b35e73ff1027e36949073efc0f36f", + "5598a5fe92b467882d3a4e316c2ac107", + "f85f5510d85c3d8597a4b8a3fc177afa", + "50f704f5c9b607dded1292f4f7629096" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "77649db0d754977f491027ad553aafec7d3f94252764508aad545d26f39c3926", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "11e33a2095244285ad1ae8d000034bf2", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1e6654d17d454951848ea2e61337a186", + "2310e88aa2a541b4a3a573100f28dcad", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + } + } + ], + "gentle_01234": [ + { + "train_samples_baselines": [ + "11e33a2095244285ad1ae8d000034bf2", + "0dda565c44d35ddc944cae93e806df1d", + "6783ef519fdc4d7bb71f66db00c264e7", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "a17a6c0b50f444b380cd3867d55fee17", + "0dda565c44d35ddc944cae93e806df1d", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "a17a6c0b50f444b380cd3867d55fee17", + "e386c1a131f5492f966c7fd71a723ae9", + "e386c1a131f5492f966c7fd71a723ae9", + "6783ef519fdc4d7bb71f66db00c264e7", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "11e33a2095244285ad1ae8d000034bf2", + "bb81df685f9a58e59ecb171a9cceb33e", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "e386c1a131f5492f966c7fd71a723ae9", + "a17a6c0b50f444b380cd3867d55fee17" + ], + "train_samples": [ + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "245d11d70caf412eb50b80db74cfb5a0", + "35428ae77661442ab6f33894965097ef", + "3a901bfee163599cb81f0751418e5ece", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "4d8fb279215248f8acbfad4f5e395b2e", + "5990d64f12064184ac3587e5004bac94", + "5f6f4011eba24b46a09cac472f54fff4", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "6be5dc20d4365c68ae8ef3bae8a17615", + "a04e965c08014d239d8b8a393e1a3cdd", + "b8b64d9c72a35d9486886e9b726261cf", + "cc080755498055de93cebb72440a86c6", + "e386c1a131f5492f966c7fd71a723ae9", + "f9bf7b07efb646ecbc9d989085c45da8" + ], + "train_samples_hashes": [ + "82d6de46d567d7269e33254ae97e5438", + "698196ffa010965e77369f23d25dee13", + "f68a5f99d2430789417bfec590c86048", + "dc6ab5bb524c1b40c954247a4f33e6c5", + "537013e33ad9f4cafe86fc16f07c6bad", + "5a4e3bfbfe29a777450556dee062d92f", + "680349aa8ffea35b1f3bc0174b398a0d", + "11c88f49a40db5a915e37952eb0d7053", + "9979a4c1ee404b9a914673b8429008de", + "378ac60f3f79eac1e573015cfe1d5d38", + "26ae001a83106a1ff76cc174e0797c70", + "6313b22ddcd216e342cf188494d40bbc", + "75b4ec8c7754d1815c1a0a1c4566827a", + "3572dd7ed11afe351f00c15cb16e62f1", + "fd447dc95cdda67db7991144d161c076", + "8e401738fdbc28eb83c176ceb4cad975", + "92850292fd8bd6a522c476d1ce6b14d7", + "df68d53d89636da356cfa7e64eb41e90" + ], + "test_samples_baselines": [ + "c7647d63ffa142a181f3bc6d78795879", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "8549e70b7f975ae68b059707d6673904", + "f12e13839aa35418b517ce7c64e0da71", + "934a4597164b46a7a2728e1ebf583b32", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "3828621236495616b794ce84cf8c84ea", + "779261e592034cf882881c2f70a20728", + "bbe78dcf65cf4990b574beb1df877cfc" + ], + "test_samples": [ + "2888456b9ece4be3bd11a8cc7dbf349b", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3731111840514a3a84fa9d182c7fb512", + "3f14e6ab62885cecad3876e7b3caf6dd", + "4f60288e7a90582a841638ff2aaca5cc", + "5838d4e3197541219e46a399693b4d50", + "68c17889e0f146f7a0772ecd1c0e63cf", + "acd65eb4f78a5539a10bb7f268c00b8b", + "e5d0ca629a2e49bcb76945a581cd1a98", + "ebb3d1ed8560449a8e1113f5f0aee8da" + ], + "test_samples_hashes": [ + "2b7e346a96d4715c8f02067c623fdc62", + "243f64406b1770defea4dfd9855d4090", + "d70d86fcdb1ed345b6cf75172957e4f7", + "f6112bf6818c85259ec05f4aef8013d6", + "843f2e5668569c0c4c2a29931dc3b2a9", + "83cc529948817b70b62e5338cac2f443", + "3681ea0983bf36d9744aa3516845b1b1", + "aa813762880b0c48ade181b6ffc371fb", + "7affebd7a60baafa0c4cd5e84b9ef2ef", + "76f91b400bf19af2e7734c0a010fe808" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "c4182e71c46d98f15d8f3c3a1ea329d76b3e013dfc0a4aad269d58f36f80e462", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35ce9600f2b649f39794bad12c8451f5", + "3828621236495616b794ce84cf8c84ea", + "3b00a79d1fff4b02a8566c05ad618150", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "55cd6c88ee704897a0318d3680135fb7", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "6407da2f934a51c5951cb7161ace6674", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "b248181b23624fdabde51fb023f87d70", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e4cf1fe8d4165cb1845b624096980de5", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "0272b7952029539d9bd40ce24d621991", + "3828621236495616b794ce84cf8c84ea", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "0272b7952029539d9bd40ce24d621991", + "f12e13839aa35418b517ce7c64e0da71", + "bb81df685f9a58e59ecb171a9cceb33e", + "3828621236495616b794ce84cf8c84ea", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "bbe78dcf65cf4990b574beb1df877cfc", + "bb81df685f9a58e59ecb171a9cceb33e", + "0dda565c44d35ddc944cae93e806df1d", + "f12e13839aa35418b517ce7c64e0da71", + "f12e13839aa35418b517ce7c64e0da71", + "c7647d63ffa142a181f3bc6d78795879", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035" + ], + "train_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "148cb65464985cf09acc84595e7a7ae3", + "1e6654d17d454951848ea2e61337a186", + "25929fbc910e457d9143e2899419263b", + "2736666ef08051b18037717bad48edcc", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "5ed1c131e780539b84a3e3c3e76d73d4", + "855c0b4b25c947ddada661a7b2138f42", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "c131520e5f80470ebb16965a0c6ba8f4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "da754a2fc8be40ee99add59263424786", + "ed3abb63ca7e49dca5dbc6f41691f421" + ], + "train_samples_hashes": [ + "c61f31bc38e3b482158addeef8ba2e75", + "b443c971956232fc19386c6fc839e29f", + "01700931d35a6e95b56e0c2f17f9d088", + "b8b3a9bced67d7217431b9939f2a9078", + "a97fdbdc4907bef0fa4d576e91eafdc0", + "04d9dc6e6bff3d76299309521aae163a", + "da810cf9ce67bbdc66427b5a54425d8d", + "9c9d519dabd56d172e09e05754193d47", + "6fe2b7c42d4485213cf19077ee205163", + "c37c7d430d6417a488dc85bcf8acc26b", + "59667920e886f04b97b7074efbe2121b", + "e9eca74cd5d96b1881f09c2085afb794", + "63687f5fc9371aefd313ed10992bd199", + "8eec429a21e528fd576f160066e454dc", + "829452e9de1c781c9cd4e755f45b1cdb", + "e6d85212838c48e229c1ed574fb84361", + "1781e6f13a65e0fd53f58dcf11718fa9", + "9dbdbdef37ea51a4f6973c98511a5417" + ], + "test_samples_baselines": [ + "779261e592034cf882881c2f70a20728", + "934a4597164b46a7a2728e1ebf583b32", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "e386c1a131f5492f966c7fd71a723ae9", + "35ce9600f2b649f39794bad12c8451f5", + "8549e70b7f975ae68b059707d6673904", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "6783ef519fdc4d7bb71f66db00c264e7", + "7379873992335258a1c5571e0f7902db" + ], + "test_samples": [ + "324f1712b21d47888e27b3d794abc71b", + "3b00a79d1fff4b02a8566c05ad618150", + "4d8fb279215248f8acbfad4f5e395b2e", + "5f6f4011eba24b46a09cac472f54fff4", + "81064a33c383447293b0d82c15f891bb", + "8549e70b7f975ae68b059707d6673904", + "d31bac7398034409ad2a8d05cf5ea3f8", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3" + ], + "test_samples_hashes": [ + "635fe28f1f1a58be42225a9b61a1d56c", + "a291f1aabdec43557934be734a6a69ce", + "11c88f49a40db5a915e37952eb0d7053", + "378ac60f3f79eac1e573015cfe1d5d38", + "a25869b63aeea4d24d4b36d2ffb62d10", + "5f3821630bc45ce5aa5333059a160167", + "1e71612c7615089a03357efd6bf8ec9f", + "0f603778fccc9f44ee65fb6bf8140069", + "6028a70fcdc33ddaf14ddd1e660f34d3", + "21262e359a3b624c46be559d7db5aae5" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "be90ad7b0a702e35a642422ad8f69c8eefb5f107015263e21ab7e54d940f734c", + "other_samples": [ + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2f760a132640578b96960b7cee8ae277", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c32fcada64fe5036a1a843f7f423f6b4", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.3, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "e386c1a131f5492f966c7fd71a723ae9", + "e386c1a131f5492f966c7fd71a723ae9", + "c7647d63ffa142a181f3bc6d78795879", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3828621236495616b794ce84cf8c84ea", + "3828621236495616b794ce84cf8c84ea", + "6783ef519fdc4d7bb71f66db00c264e7", + "7379873992335258a1c5571e0f7902db", + "c7647d63ffa142a181f3bc6d78795879", + "0272b7952029539d9bd40ce24d621991", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "3828621236495616b794ce84cf8c84ea", + "e386c1a131f5492f966c7fd71a723ae9", + "7379873992335258a1c5571e0f7902db", + "7379873992335258a1c5571e0f7902db" + ], + "train_samples": [ + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "29a0305a7d8947b0983ab26ff8f5b47a", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3828621236495616b794ce84cf8c84ea", + "6407da2f934a51c5951cb7161ace6674", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "7379873992335258a1c5571e0f7902db", + "78553d66439e5f678fe0b5574efa7384", + "8399e56763da5bc59ef9f9c4b652eeee", + "866e78a3aee946dfaab06893eac3d553", + "9584da5231b4512188b5089c9a0dbc05", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "acd65eb4f78a5539a10bb7f268c00b8b", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "train_samples_hashes": [ + "365ec0182691f380b069383e4df9861e", + "841e717955365a497a1f405c126bd395", + "fd7917ec7445186cf3174db36868cb62", + "243f64406b1770defea4dfd9855d4090", + "646d80e73dfb4473643ec711415ec877", + "71151601bc29c119935779b5ae768f14", + "6a0550716d5eefc02e5a3d0ba2a08589", + "1f1a2fb77d47733588a417e23660ac4f", + "a0c2a34012ee1c36abfc33b74e72aea0", + "5b3926fe15045159c0fef5628c86baee", + "f658b04327b8b0123de2350dc1f9fe23", + "0c472fe8fc6a82d3c8973b210920a1f5", + "a4e333a6de8f677270489817573748c9", + "a362aefef1a80b9fd63926fbb535ce10", + "aa813762880b0c48ade181b6ffc371fb", + "a682076279644b3c16f136b1e48eef3a", + "21262e359a3b624c46be559d7db5aae5", + "9855205bcbf0bf2f7274ab7b17974d8e" + ], + "test_samples_baselines": [ + "f12e13839aa35418b517ce7c64e0da71", + "35ce9600f2b649f39794bad12c8451f5", + "934a4597164b46a7a2728e1ebf583b32", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "bbe78dcf65cf4990b574beb1df877cfc", + "bb81df685f9a58e59ecb171a9cceb33e", + "11e33a2095244285ad1ae8d000034bf2", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "779261e592034cf882881c2f70a20728", + "8549e70b7f975ae68b059707d6673904" + ], + "test_samples": [ + "062a37da2b3c523ea65fe50198329901", + "35ce9600f2b649f39794bad12c8451f5", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "96f1621029835fd49654f3249386c245", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a04e965c08014d239d8b8a393e1a3cdd", + "c5b764da8f9442ebbce3b0859ed481a1", + "ca8e6694b045476e8d251f55f5445ab3", + "e4cf1fe8d4165cb1845b624096980de5" + ], + "test_samples_hashes": [ + "e25f2297ed37ce231a36f6642322d064", + "b054165364e7ab8827bbcaf351c56534", + "795785f4ba2c82450626bf9385860036", + "5d776197ac4eab709532b359e5c94428", + "e9eca74cd5d96b1881f09c2085afb794", + "63687f5fc9371aefd313ed10992bd199", + "3572dd7ed11afe351f00c15cb16e62f1", + "ed4fb51fbbe9085bcc26ab5ec9f832a4", + "df4762fc4837017b44d02f3d9712953f", + "149c92837125f8dc1559cabfcc6f11d2" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.1, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "b52e60ee00536897ae764501cb148fd5fd30148339a341e72bb4f61920fee41d", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "105de8c7a17f4728931344dd36a28313", + "11e33a2095244285ad1ae8d000034bf2", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1b364086c3c740789d832cfebdd6cb1e", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "3731111840514a3a84fa9d182c7fb512", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6ca1f93c5f9d59669a10db584518a5da", + "73e57408da7a53bba190eae95592a58b", + "779261e592034cf882881c2f70a20728", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "9820ec472b7e4371a4d40be054b4c6ae", + "a71f474e41464187af46ccbd7222cae8", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "bbe78dcf65cf4990b574beb1df877cfc", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e957c95514144e96beeead74e1ca010a", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "11e33a2095244285ad1ae8d000034bf2", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "779261e592034cf882881c2f70a20728", + "779261e592034cf882881c2f70a20728", + "a17a6c0b50f444b380cd3867d55fee17", + "6783ef519fdc4d7bb71f66db00c264e7", + "f12e13839aa35418b517ce7c64e0da71", + "8549e70b7f975ae68b059707d6673904", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "8549e70b7f975ae68b059707d6673904", + "7379873992335258a1c5571e0f7902db", + "779261e592034cf882881c2f70a20728", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "e386c1a131f5492f966c7fd71a723ae9", + "779261e592034cf882881c2f70a20728" + ], + "train_samples": [ + "11e33a2095244285ad1ae8d000034bf2", + "1b364086c3c740789d832cfebdd6cb1e", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "324f1712b21d47888e27b3d794abc71b", + "3fd9c7f94e355f65bada44cc585647d6", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "5af136001c8e55ed96639fb169b820df", + "73e57408da7a53bba190eae95592a58b", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "9820ec472b7e4371a4d40be054b4c6ae", + "cc34c415d1465097ae993175b76900b8", + "da62c8d5d796459081533a11ff95d0c9", + "e6614388019a4d22bdd9140403a47547", + "e957c95514144e96beeead74e1ca010a", + "fab4f5f11b874ef58e909995690d57ec" + ], + "train_samples_hashes": [ + "9199d41302f5567aed152b1f3201db49", + "3fca2198a402cddcc08c6f711ddd1ba5", + "1f67224e94945758df3fb47141ef3535", + "635fe28f1f1a58be42225a9b61a1d56c", + "ee0fc1ebb0b9837152c273192bf66319", + "64f3ddce27f31b9286f3e236b7aec0fe", + "843f2e5668569c0c4c2a29931dc3b2a9", + "b339372c4d697c6866bfa02ddb77fd53", + "9dad8771ba3f0ce58f7a85d79f7af2d6", + "9eff31f3977f47d94aee7b8d1c72570c", + "6adebb0e7c5af169480e03a11d7cab26", + "5f3821630bc45ce5aa5333059a160167", + "8b8b3520edb16081225b5692f78e7127", + "a29268334ba4a9303e4362444dee0979", + "b10e3c032bcc3535b1bed71322c70182", + "0f603778fccc9f44ee65fb6bf8140069", + "34c00222d6beb05c3f71a4a79887107e", + "bad3ce5eb8272ff3ff26555f1ef53db7" + ], + "test_samples_baselines": [ + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "3828621236495616b794ce84cf8c84ea", + "0272b7952029539d9bd40ce24d621991", + "0dda565c44d35ddc944cae93e806df1d", + "bbe78dcf65cf4990b574beb1df877cfc", + "35ce9600f2b649f39794bad12c8451f5", + "6d6e5f8faa79553cb3b6e6f4c31d4de2", + "934a4597164b46a7a2728e1ebf583b32", + "bb81df685f9a58e59ecb171a9cceb33e", + "c7647d63ffa142a181f3bc6d78795879" + ], + "test_samples": [ + "105dab6037814698b8c5fb1b7cb0d358", + "2736666ef08051b18037717bad48edcc", + "2f760a132640578b96960b7cee8ae277", + "42a04eb48f2359898f15b1581dab6727", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "bb81df685f9a58e59ecb171a9cceb33e", + "d67bbb7d87d745ad8b5f8871e4ae9a10" + ], + "test_samples_hashes": [ + "b8b3a9bced67d7217431b9939f2a9078", + "9c9d519dabd56d172e09e05754193d47", + "df46b10959e0653936736c9ad43bbfe1", + "09d201c2304591700da298bc6e543570", + "035d38b8be32e0e01aab8449ac8c23d3", + "f9455494e70645b8c2ea7aba858e9449", + "a362aefef1a80b9fd63926fbb535ce10", + "e5c8469aedae7cbf8b54337f754e5ad4", + "402af21e1c849e04ac412fe2d544fbf5", + "75de90efb4b12116d19217bae9278b00" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "2a4aad740cda8db3c743d3785955861791ad44e6fb487e6459bf888dc1c9267f", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105de8c7a17f4728931344dd36a28313", + "1276239e6f2a43c58fe7ab267bd41d94", + "12dd1c92dc4649d4808421b7c2dbb124", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1aac2ed4a25544c894c596067aeb18ae", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "2310e88aa2a541b4a3a573100f28dcad", + "245d11d70caf412eb50b80db74cfb5a0", + "25929fbc910e457d9143e2899419263b", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2afe39e081a14923a1d46c052479dc3e", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3b00a79d1fff4b02a8566c05ad618150", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3ff3d291e19c586a958adb1f94864c17", + "4483b5b67b3f4d9abae899f3bea83fa0", + "45aa02ea71c0510aae5f198aef30fef5", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "4bdd20ee31ce447eb8b038067bdd66d7", + "4d8fb279215248f8acbfad4f5e395b2e", + "55cd6c88ee704897a0318d3680135fb7", + "5838d4e3197541219e46a399693b4d50", + "5990d64f12064184ac3587e5004bac94", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "6ca1f93c5f9d59669a10db584518a5da", + "7379873992335258a1c5571e0f7902db", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "779261e592034cf882881c2f70a20728", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9584da5231b4512188b5089c9a0dbc05", + "95ed9fef81a25479881312e047965573", + "9698d8207595571e9187650324239bb0", + "96f1621029835fd49654f3249386c245", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b248181b23624fdabde51fb023f87d70", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bbe78dcf65cf4990b574beb1df877cfc", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "be961034c82352e09e4c5952be16637a", + "c131520e5f80470ebb16965a0c6ba8f4", + "c32fcada64fe5036a1a843f7f423f6b4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "d107128850bd502784cdb511d7efa744", + "d217e2d448755e4bb51130635edf22b9", + "d31bac7398034409ad2a8d05cf5ea3f8", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f3359983559e491da7139fb1d378f667", + "f9bf7b07efb646ecbc9d989085c45da8", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.4, + "euclidean_centroid": 0.2, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [ + "bbe78dcf65cf4990b574beb1df877cfc", + "934a4597164b46a7a2728e1ebf583b32", + "bbe78dcf65cf4990b574beb1df877cfc", + "934a4597164b46a7a2728e1ebf583b32", + "3828621236495616b794ce84cf8c84ea", + "934a4597164b46a7a2728e1ebf583b32", + "35ce9600f2b649f39794bad12c8451f5", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "934a4597164b46a7a2728e1ebf583b32", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "7379873992335258a1c5571e0f7902db", + "bbe78dcf65cf4990b574beb1df877cfc", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "bb81df685f9a58e59ecb171a9cceb33e", + "35ce9600f2b649f39794bad12c8451f5", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "bbe78dcf65cf4990b574beb1df877cfc" + ], + "train_samples": [ + "12dd1c92dc4649d4808421b7c2dbb124", + "2310e88aa2a541b4a3a573100f28dcad", + "2afe39e081a14923a1d46c052479dc3e", + "3b00a79d1fff4b02a8566c05ad618150", + "461103bd3ac255b2a1e6bd5a4122cbc4", + "5838d4e3197541219e46a399693b4d50", + "7b377fd33ec55476b4dd35ce0f6adf5b", + "8f80071112174821a3595067c2bf1537", + "90872800e402540eb77c39d26a123b23", + "934a4597164b46a7a2728e1ebf583b32", + "9698d8207595571e9187650324239bb0", + "b248181b23624fdabde51fb023f87d70", + "bbe78dcf65cf4990b574beb1df877cfc", + "be961034c82352e09e4c5952be16637a", + "c32fcada64fe5036a1a843f7f423f6b4", + "d217e2d448755e4bb51130635edf22b9", + "e9cc36d951655bdfbcf90d0bb2322a0e", + "f3359983559e491da7139fb1d378f667" + ], + "train_samples_hashes": [ + "671dc61016016609cc90760b4d2b9013", + "7e6b89b42854d0c6b9dd6dabdb037bda", + "484e10d4b2e22c945701c5abb5e96c53", + "a291f1aabdec43557934be734a6a69ce", + "1998992a9f55b8529d050692671866b4", + "83cc529948817b70b62e5338cac2f443", + "f9455494e70645b8c2ea7aba858e9449", + "0b8ddbea3b36ee43473118eb56d7ae9a", + "fda06dc08847de9fb0c4a88b7350dfc7", + "a75236ebbaa2deb369c720d5a860c93f", + "c554c189cfd1f44941d4ad41c2d8e289", + "bc02b76c7cd403f0fca3090e321f8414", + "09897ffa321559f9ad73838b55cbba8d", + "5e6f000fa3912ff2d686d24bef62f9b6", + "8e4b8dd31d946bf8483cb98cc6ca28f4", + "a711d5ad4b1191cbb8bf4e986e311740", + "559649d78778f524144fd61b98183086", + "b7877b6e765cb0b52a88931e79f15116" + ], + "test_samples_baselines": [ + "1032bb73c0b55d1ba0c4685bb20f08f6", + "8549e70b7f975ae68b059707d6673904", + "e386c1a131f5492f966c7fd71a723ae9", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "6783ef519fdc4d7bb71f66db00c264e7", + "a17a6c0b50f444b380cd3867d55fee17", + "c7647d63ffa142a181f3bc6d78795879", + "0dda565c44d35ddc944cae93e806df1d", + "779261e592034cf882881c2f70a20728", + "2c5b25ed588c52f5a5b4f58fc232e4bb" + ], + "test_samples": [ + "105de8c7a17f4728931344dd36a28313", + "1276239e6f2a43c58fe7ab267bd41d94", + "1aac2ed4a25544c894c596067aeb18ae", + "1b364086c3c740789d832cfebdd6cb1e", + "245d11d70caf412eb50b80db74cfb5a0", + "4483b5b67b3f4d9abae899f3bea83fa0", + "4bdd20ee31ce447eb8b038067bdd66d7", + "6ca1f93c5f9d59669a10db584518a5da", + "779261e592034cf882881c2f70a20728", + "95ed9fef81a25479881312e047965573" + ], + "test_samples_hashes": [ + "1e55b3e782900756be811ce23aa9c293", + "7a563abc0a8b48d95b01f5708a0c44cd", + "841e717955365a497a1f405c126bd395", + "3fca2198a402cddcc08c6f711ddd1ba5", + "f68a5f99d2430789417bfec590c86048", + "5a4e3bfbfe29a777450556dee062d92f", + "c61b35e73ff1027e36949073efc0f36f", + "5598a5fe92b467882d3a4e316c2ac107", + "f85f5510d85c3d8597a4b8a3fc177afa", + "50f704f5c9b607dded1292f4f7629096" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "371820c4b19e3a37da9b26b859a20d0d643339ce9e075b872762b77453d0240d", + "other_samples": [ + "0272b7952029539d9bd40ce24d621991", + "0310deb38f465db18ad47c098129e23d", + "062a37da2b3c523ea65fe50198329901", + "07f55a16d69449fcb096b003f6593691", + "0dda565c44d35ddc944cae93e806df1d", + "1032bb73c0b55d1ba0c4685bb20f08f6", + "105dab6037814698b8c5fb1b7cb0d358", + "11e33a2095244285ad1ae8d000034bf2", + "148cb65464985cf09acc84595e7a7ae3", + "1a6d28e85a7a41a292e5c95ea91ee2cb", + "1e6654d17d454951848ea2e61337a186", + "22bd886147e847e9a0420134bbfcd2c8", + "25929fbc910e457d9143e2899419263b", + "266ff8b228ea4c2bbc98b40b4b29ae1e", + "2736666ef08051b18037717bad48edcc", + "2888456b9ece4be3bd11a8cc7dbf349b", + "29a0305a7d8947b0983ab26ff8f5b47a", + "2c5b25ed588c52f5a5b4f58fc232e4bb", + "2f760a132640578b96960b7cee8ae277", + "324f1712b21d47888e27b3d794abc71b", + "35428ae77661442ab6f33894965097ef", + "354dbe3c5b4b4ef29b2e2a1a0ae5c035", + "35ce9600f2b649f39794bad12c8451f5", + "3731111840514a3a84fa9d182c7fb512", + "3828621236495616b794ce84cf8c84ea", + "3a901bfee163599cb81f0751418e5ece", + "3f14e6ab62885cecad3876e7b3caf6dd", + "3fd9c7f94e355f65bada44cc585647d6", + "3ff3d291e19c586a958adb1f94864c17", + "42a04eb48f2359898f15b1581dab6727", + "45aa02ea71c0510aae5f198aef30fef5", + "4d8fb279215248f8acbfad4f5e395b2e", + "4e53c84aa5024bf4ae26bb23d7a4d07e", + "4f26efbd7d274f7fa043588a6d4e7204", + "4f60288e7a90582a841638ff2aaca5cc", + "55cd6c88ee704897a0318d3680135fb7", + "5990d64f12064184ac3587e5004bac94", + "5af136001c8e55ed96639fb169b820df", + "5ed1c131e780539b84a3e3c3e76d73d4", + "5f6f4011eba24b46a09cac472f54fff4", + "6407da2f934a51c5951cb7161ace6674", + "664fa7515eea439692b7f460afbf1631", + "6783ef519fdc4d7bb71f66db00c264e7", + "68c17889e0f146f7a0772ecd1c0e63cf", + "6be5dc20d4365c68ae8ef3bae8a17615", + "6c4ff7108da34c2d94ba2bb300dd1ee7", + "7379873992335258a1c5571e0f7902db", + "73e57408da7a53bba190eae95592a58b", + "74b7a23252a94ac4aa7aa071d7b17aa5", + "78553d66439e5f678fe0b5574efa7384", + "7b2b2d6cce1c444d8fc6bc0be9ac42ee", + "7dacaf46a9ec45cf9c7e1aabad31e83f", + "81064a33c383447293b0d82c15f891bb", + "8399e56763da5bc59ef9f9c4b652eeee", + "84bbd6c06ddb4b54a3f12c882c2eccc4", + "8549e70b7f975ae68b059707d6673904", + "855c0b4b25c947ddada661a7b2138f42", + "866e78a3aee946dfaab06893eac3d553", + "9584da5231b4512188b5089c9a0dbc05", + "96f1621029835fd49654f3249386c245", + "9820ec472b7e4371a4d40be054b4c6ae", + "98ce3759ae534c9f8ee938e151b01032", + "9a03fccac5e1443fbea77e7243b73b84", + "a0399f67cb5f4165a4dc00fa72104bb6", + "a04e965c08014d239d8b8a393e1a3cdd", + "a3d4d1a02f6f4d74a743b6f0212ec691", + "a71f474e41464187af46ccbd7222cae8", + "acd65eb4f78a5539a10bb7f268c00b8b", + "b8b64d9c72a35d9486886e9b726261cf", + "ba1a51d50f0b5031b0f10f2b48f8f5b5", + "bb81df685f9a58e59ecb171a9cceb33e", + "be12a2ede7744ecdae83f9e0d1f2a47f", + "c131520e5f80470ebb16965a0c6ba8f4", + "c3b3ddcba6ff5526bd21d1d06ee66506", + "c5b764da8f9442ebbce3b0859ed481a1", + "c67fd00576b04ac2bb3baf818c034ae4", + "c7647d63ffa142a181f3bc6d78795879", + "ca8e6694b045476e8d251f55f5445ab3", + "cc080755498055de93cebb72440a86c6", + "cc34c415d1465097ae993175b76900b8", + "d107128850bd502784cdb511d7efa744", + "d31bac7398034409ad2a8d05cf5ea3f8", + "d67bbb7d87d745ad8b5f8871e4ae9a10", + "da62c8d5d796459081533a11ff95d0c9", + "da754a2fc8be40ee99add59263424786", + "e386c1a131f5492f966c7fd71a723ae9", + "e4cf1fe8d4165cb1845b624096980de5", + "e5d0ca629a2e49bcb76945a581cd1a98", + "e6614388019a4d22bdd9140403a47547", + "e6b5a1b16cc15dbe8d1fe3b51eb03100", + "e755a27e73ee5af2bf1337a1a71d0ea3", + "e957c95514144e96beeead74e1ca010a", + "ebb3d1ed8560449a8e1113f5f0aee8da", + "ed3abb63ca7e49dca5dbc6f41691f421", + "f9bf7b07efb646ecbc9d989085c45da8", + "fab4f5f11b874ef58e909995690d57ec", + "fc653cb5b3c05a259ddecd3547667a15" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + } + } + ] +} \ No newline at end of file diff --git a/questions/MassSlide/dataframes/00af86611bf0452986e018425db6cdd2.parquet b/questions/MassSlide/dataframes/00af86611bf0452986e018425db6cdd2.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ad950b000e389f215e4e65897f801f7134ea3f67 --- /dev/null +++ b/questions/MassSlide/dataframes/00af86611bf0452986e018425db6cdd2.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4638a2b71a0ba965a1be87285f512ea746195291d2b30ab54196c98831f46398 +size 7406 diff --git a/questions/MassSlide/dataframes/03cf9a5fd2af43aa8b2413596e24ffa4.parquet b/questions/MassSlide/dataframes/03cf9a5fd2af43aa8b2413596e24ffa4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7d60586f7a2ace9907c783660cbd021b0eded89e --- /dev/null +++ b/questions/MassSlide/dataframes/03cf9a5fd2af43aa8b2413596e24ffa4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aac07117e3536e60710ef78c6b9181c9b09ec0ae2f92e7c4d15a3be98f59aa79 +size 5847 diff --git a/questions/MassSlide/dataframes/0655d91288195e93a7903500183734dd.parquet b/questions/MassSlide/dataframes/0655d91288195e93a7903500183734dd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..536891d34eface5115a7459981e04070ad5c7e36 --- /dev/null +++ b/questions/MassSlide/dataframes/0655d91288195e93a7903500183734dd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9806cb38e6ded917fdcfd68be99728060434d71b9ec2221f57aab61092f11c62 +size 7974 diff --git a/questions/MassSlide/dataframes/07fd7c347c7354bbb738d7765a2fea14.parquet b/questions/MassSlide/dataframes/07fd7c347c7354bbb738d7765a2fea14.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b6a6a2e64467e55c5e451a99640601996e517eb7 --- /dev/null +++ b/questions/MassSlide/dataframes/07fd7c347c7354bbb738d7765a2fea14.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a89a17266b6f91fa3ee6a4a0154fdb16208fa73f5525cec15524fdf8151cfa9 +size 7974 diff --git a/questions/MassSlide/dataframes/0982071fbb684307913a380272a1ae91.parquet b/questions/MassSlide/dataframes/0982071fbb684307913a380272a1ae91.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4bfa3ea895febb18576d300667c59b418a20f0be --- /dev/null +++ b/questions/MassSlide/dataframes/0982071fbb684307913a380272a1ae91.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ef3a5d7a15f64314c1e4de7cb52cb211be4ad1bd26fa437891a146e189e1c15 +size 6276 diff --git a/questions/MassSlide/dataframes/0c5a280f084b4d1fa52dddd30269c1d4.parquet b/questions/MassSlide/dataframes/0c5a280f084b4d1fa52dddd30269c1d4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1482bd4be92b833fa3ca0034f50963e60a1454d2 --- /dev/null +++ b/questions/MassSlide/dataframes/0c5a280f084b4d1fa52dddd30269c1d4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a490ea84234221623e0f29b36bdd8a9e5505a2f45ec78dff23839af67a61f27 +size 6506 diff --git a/questions/MassSlide/dataframes/0d79edc07243565d8d8b5380ef81e340.parquet b/questions/MassSlide/dataframes/0d79edc07243565d8d8b5380ef81e340.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0528da9e0855212311d4183e6050f9665e991423 --- /dev/null +++ b/questions/MassSlide/dataframes/0d79edc07243565d8d8b5380ef81e340.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76f6a23e07d7a39722aa82e854fdaff9797f5da468f91d41dd9d80bea2adf43f +size 7981 diff --git a/questions/MassSlide/dataframes/0dc61f09ac3a57cba3e986abd378a58c.parquet b/questions/MassSlide/dataframes/0dc61f09ac3a57cba3e986abd378a58c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4bd08283d13fb44af0a0b57e6379896dc0310fe0 --- /dev/null +++ b/questions/MassSlide/dataframes/0dc61f09ac3a57cba3e986abd378a58c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8fe14b52d2cd394304fd35b47afc2a2f5330954dfe23f9220ec97d789c784c7 +size 7967 diff --git a/questions/MassSlide/dataframes/115036fcb4cb40278192aeaec3c948ba.parquet b/questions/MassSlide/dataframes/115036fcb4cb40278192aeaec3c948ba.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2726f32a7679650bf6cff4176c4de74191ad6a1a --- /dev/null +++ b/questions/MassSlide/dataframes/115036fcb4cb40278192aeaec3c948ba.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5340eb5a43000fd262273bacbdbfa1c8a095691301b8bf285b2ee9cfcf92faeb +size 5712 diff --git a/questions/MassSlide/dataframes/1d716244585943cc9e58b841706c0a36.parquet b/questions/MassSlide/dataframes/1d716244585943cc9e58b841706c0a36.parquet new file mode 100644 index 0000000000000000000000000000000000000000..65d137f50cc2dcaebbd1931bc054182c61b4b7cc --- /dev/null +++ b/questions/MassSlide/dataframes/1d716244585943cc9e58b841706c0a36.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c9796fd87cf7d2c996f091421688ef032d816836365138b9745f3167584ac96 +size 7981 diff --git a/questions/MassSlide/dataframes/1dbde2fdb75b5ca294baa80857327b67.parquet b/questions/MassSlide/dataframes/1dbde2fdb75b5ca294baa80857327b67.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a58329d1089da6ebdf96e34790a9f9aab053c4e8 --- /dev/null +++ b/questions/MassSlide/dataframes/1dbde2fdb75b5ca294baa80857327b67.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5becfdb46c7ee4a1afef8fc9ebb38592e87dee583bdf6872918fde0d5783f1aa +size 6013 diff --git a/questions/MassSlide/dataframes/1f0a940c521643948cd6a4a3f8c15067.parquet b/questions/MassSlide/dataframes/1f0a940c521643948cd6a4a3f8c15067.parquet new file mode 100644 index 0000000000000000000000000000000000000000..72753dc0f39cea2a556057bbb805c9e57686010e --- /dev/null +++ b/questions/MassSlide/dataframes/1f0a940c521643948cd6a4a3f8c15067.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf2e166fa919a6efcb13bd67fcef0cf9810fc985e7e735c95aed7bcc215ba51a +size 6236 diff --git a/questions/MassSlide/dataframes/208ef509e8b04166ae5333a2877f1295.parquet b/questions/MassSlide/dataframes/208ef509e8b04166ae5333a2877f1295.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1a67046b946c574a669af0b78f307fb218b3bb43 --- /dev/null +++ b/questions/MassSlide/dataframes/208ef509e8b04166ae5333a2877f1295.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48d5c14ba6f65fb417d3785688ac7e469dd7f30a2c7dfb8abf73be17ad46b53e +size 7899 diff --git a/questions/MassSlide/dataframes/223c934b74a24e5f9df19997035eef2f.parquet b/questions/MassSlide/dataframes/223c934b74a24e5f9df19997035eef2f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f3964165db7e2502ead7be6c0ea81ef1fddf4b19 --- /dev/null +++ b/questions/MassSlide/dataframes/223c934b74a24e5f9df19997035eef2f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6ed709bb27ac9bd26ecca37e93bdd46e21ddb5af6e7d38a58449a497a9665f8 +size 7981 diff --git a/questions/MassSlide/dataframes/23cca9a4b30252e89b56e757b3b6b0f0.parquet b/questions/MassSlide/dataframes/23cca9a4b30252e89b56e757b3b6b0f0.parquet new file mode 100644 index 0000000000000000000000000000000000000000..378707e1d05ba5c99109572d96d506f45abfa857 --- /dev/null +++ b/questions/MassSlide/dataframes/23cca9a4b30252e89b56e757b3b6b0f0.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4eaadb80828868d64c691b7dbf94f903e811c7a032fbef9eadc817b63ccb5caf +size 5531 diff --git a/questions/MassSlide/dataframes/260750f9d27a4578ba899fcfb02ac653.parquet b/questions/MassSlide/dataframes/260750f9d27a4578ba899fcfb02ac653.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0b4bdfbe51454f93ab91f0642302643aae92bb9d --- /dev/null +++ b/questions/MassSlide/dataframes/260750f9d27a4578ba899fcfb02ac653.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48bb163bcd825fec79038308fff104eed228138a9271f148c1fe273370c85220 +size 6652 diff --git a/questions/MassSlide/dataframes/2713905d0dfa4d56aac01b96531ab3e7.parquet b/questions/MassSlide/dataframes/2713905d0dfa4d56aac01b96531ab3e7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..de614ea6bf84da920a224cba51be82b36c254661 --- /dev/null +++ b/questions/MassSlide/dataframes/2713905d0dfa4d56aac01b96531ab3e7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d7003ba84ed45e835a33da34393bf80dd31b47ae63440ab67bd27d0fcdb381e +size 7967 diff --git a/questions/MassSlide/dataframes/299fe66074144eeeae74735cf42d161b.parquet b/questions/MassSlide/dataframes/299fe66074144eeeae74735cf42d161b.parquet new file mode 100644 index 0000000000000000000000000000000000000000..597e618aacb69f78aac2b976a01f42061a40deee --- /dev/null +++ b/questions/MassSlide/dataframes/299fe66074144eeeae74735cf42d161b.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:72a7f799cc7ed7dbb506566af75ac68b597bfa7ed200f58bf2d5a5dde01f9429 +size 7625 diff --git a/questions/MassSlide/dataframes/2dec7910f3a25f448d279e1bcd61c763.parquet b/questions/MassSlide/dataframes/2dec7910f3a25f448d279e1bcd61c763.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c4e140ed6364b208ba63357e0681fc7d068825fc --- /dev/null +++ b/questions/MassSlide/dataframes/2dec7910f3a25f448d279e1bcd61c763.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50550de6d92536bc50bc934bb5e154df645830426d7e9e1058e773e27c146a9f +size 7973 diff --git a/questions/MassSlide/dataframes/2eaaaa5d6f8e4c1b8447b4389e52853e.parquet b/questions/MassSlide/dataframes/2eaaaa5d6f8e4c1b8447b4389e52853e.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a79757af391e05f07a64e7d878f7cb692ded0827 --- /dev/null +++ b/questions/MassSlide/dataframes/2eaaaa5d6f8e4c1b8447b4389e52853e.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2987c12b4790b67a9f7fe1c893b4bc7d8226bfa260135aef4032bc8826c13dfb +size 5970 diff --git a/questions/MassSlide/dataframes/2f00e9a90e3c4484aa48107c021745aa.parquet b/questions/MassSlide/dataframes/2f00e9a90e3c4484aa48107c021745aa.parquet new file mode 100644 index 0000000000000000000000000000000000000000..54e8c436ccfbebe59e4bf41059ad07ec339f7da0 --- /dev/null +++ b/questions/MassSlide/dataframes/2f00e9a90e3c4484aa48107c021745aa.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25b6ef7d114fb26415b10554da5b54c139f6f416b5fc41fa3cc525edddbc25d3 +size 5933 diff --git a/questions/MassSlide/dataframes/2fc73742c4235516a326bc749d661972.parquet b/questions/MassSlide/dataframes/2fc73742c4235516a326bc749d661972.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2bba0f836327a6be108dcbbe74b45a170779b3f4 --- /dev/null +++ b/questions/MassSlide/dataframes/2fc73742c4235516a326bc749d661972.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c78dbdada995e54b7734d9801c5d393ac2ee50a360a786b967b74a5371e56c37 +size 7988 diff --git a/questions/MassSlide/dataframes/322aa53bc0ae43c08d03196c3fa01dcf.parquet b/questions/MassSlide/dataframes/322aa53bc0ae43c08d03196c3fa01dcf.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e49fc2d46f1a55577bd0060970ecc7144b913b23 --- /dev/null +++ b/questions/MassSlide/dataframes/322aa53bc0ae43c08d03196c3fa01dcf.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4239f208850007f8f8e6a55ba94d788952bf412560e49ea97136e8312d02d343 +size 5813 diff --git a/questions/MassSlide/dataframes/3273b1d94d045773a3ed5d180a4511dc.parquet b/questions/MassSlide/dataframes/3273b1d94d045773a3ed5d180a4511dc.parquet new file mode 100644 index 0000000000000000000000000000000000000000..43c35be5e17da0cc682a5751f68484d29c1dfc9e --- /dev/null +++ b/questions/MassSlide/dataframes/3273b1d94d045773a3ed5d180a4511dc.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d281c71e8e5b7214f3152d5ad43e5495115f1fc4911b341fab6e939e8ab9e48e +size 5681 diff --git a/questions/MassSlide/dataframes/337efe3f652b4c12b24b70a63f25adff.parquet b/questions/MassSlide/dataframes/337efe3f652b4c12b24b70a63f25adff.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f7f492ed48f53ef754a3c0dbdc7d95ba06670131 --- /dev/null +++ b/questions/MassSlide/dataframes/337efe3f652b4c12b24b70a63f25adff.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df1404102de08f019acffa8584b71aa38cb533949f6ba1b041922ecaf99df1ca +size 5797 diff --git a/questions/MassSlide/dataframes/38482b8747bc4bfe88ac617356e9c6fe.parquet b/questions/MassSlide/dataframes/38482b8747bc4bfe88ac617356e9c6fe.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cca929f5effdbea5fd077d27587dd569899306fb --- /dev/null +++ b/questions/MassSlide/dataframes/38482b8747bc4bfe88ac617356e9c6fe.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6097531364bb485fbcea8bc9da15017a1dddc4389323031737f7367f95e9ae6 +size 7967 diff --git a/questions/MassSlide/dataframes/38e9b9cde3d547a9924b3af04070203a.parquet b/questions/MassSlide/dataframes/38e9b9cde3d547a9924b3af04070203a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ccfd84caa821158a564d175c6961506e17449d9d --- /dev/null +++ b/questions/MassSlide/dataframes/38e9b9cde3d547a9924b3af04070203a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3643a12ef9a1115b960f48348a48589c3859abd5bd4b50ce63c2997ca2acb1f2 +size 6032 diff --git a/questions/MassSlide/dataframes/3ad87b80437b43fda95e551a2efc7b87.parquet b/questions/MassSlide/dataframes/3ad87b80437b43fda95e551a2efc7b87.parquet new file mode 100644 index 0000000000000000000000000000000000000000..8d265d94ef358dba329fe098d1fddba45079b195 --- /dev/null +++ b/questions/MassSlide/dataframes/3ad87b80437b43fda95e551a2efc7b87.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76be1c819dfc945d0f4e652df0e4efd8fada83a71b2312fa0bc332bae87dd4d2 +size 6051 diff --git a/questions/MassSlide/dataframes/3ba3a08aac1441ab9f7062076ab2e050.parquet b/questions/MassSlide/dataframes/3ba3a08aac1441ab9f7062076ab2e050.parquet new file mode 100644 index 0000000000000000000000000000000000000000..1fcb4297aaed0677be835718e7a334900b154457 --- /dev/null +++ b/questions/MassSlide/dataframes/3ba3a08aac1441ab9f7062076ab2e050.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bf66bed273309aac6105e6ec5470e6aeb3fbb49664e74ba6200d0292982d7dd +size 6064 diff --git a/questions/MassSlide/dataframes/3c9308c66d644b07be5b2c3ee6ffa4e6.parquet b/questions/MassSlide/dataframes/3c9308c66d644b07be5b2c3ee6ffa4e6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d7291d0e637934cc206f11c131ed48c274e09ac9 --- /dev/null +++ b/questions/MassSlide/dataframes/3c9308c66d644b07be5b2c3ee6ffa4e6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29267059e472a9b9d6c3e21c5855b7d4ab3bf81cd4019331fad5dfc76c5286a1 +size 5770 diff --git a/questions/MassSlide/dataframes/3cdf2c46f20e55c0b081c3745fbf8c40.parquet b/questions/MassSlide/dataframes/3cdf2c46f20e55c0b081c3745fbf8c40.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6f3f41f9dd6e86aaeb141871ececf4e30ba6f5cf --- /dev/null +++ b/questions/MassSlide/dataframes/3cdf2c46f20e55c0b081c3745fbf8c40.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32d2c16d97f15cfc336726bb9ce86e584eec7571b3735e9fb59755638ad43150 +size 7974 diff --git a/questions/MassSlide/dataframes/3d97964fb7c24cc291fbc25dea19c9d4.parquet b/questions/MassSlide/dataframes/3d97964fb7c24cc291fbc25dea19c9d4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d58bfedb7df8bf4a67f2b64a94c8fb566f37f5aa --- /dev/null +++ b/questions/MassSlide/dataframes/3d97964fb7c24cc291fbc25dea19c9d4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9e1c182fbf5bb34f2304b8620a9a0c728b6fb362a982710189e334aff1db4e5 +size 7988 diff --git a/questions/MassSlide/dataframes/3dc5e2e8d7f150fbaf80df25ba9c426a.parquet b/questions/MassSlide/dataframes/3dc5e2e8d7f150fbaf80df25ba9c426a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cc5b4b3a6499ac81ec187a61eccef8aacb09b77e --- /dev/null +++ b/questions/MassSlide/dataframes/3dc5e2e8d7f150fbaf80df25ba9c426a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2f32b1688007bd72156ba31c01bee045e872d3d5a1cb78593de7671eacaae6c4 +size 5808 diff --git a/questions/MassSlide/dataframes/3dee5963901d530e9f2fde3c84a55982.parquet b/questions/MassSlide/dataframes/3dee5963901d530e9f2fde3c84a55982.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c8ddd5b0983c11088e5dbdf8f913fbbecd640663 --- /dev/null +++ b/questions/MassSlide/dataframes/3dee5963901d530e9f2fde3c84a55982.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49fd87eb42512117ffab1a3c8600e50b318aa13e2942a4aae5935f91fce72ac8 +size 5698 diff --git a/questions/MassSlide/dataframes/402307421702585c86a2ff40516deca2.parquet b/questions/MassSlide/dataframes/402307421702585c86a2ff40516deca2.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3a108280d0602f761d377e8660a3dca359e09b2f --- /dev/null +++ b/questions/MassSlide/dataframes/402307421702585c86a2ff40516deca2.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c278a152962ce2ea631bcc45c8714ecedac1765258cd0014cbbc8246be3b261c +size 7977 diff --git a/questions/MassSlide/dataframes/421875a8a7f441c4bffd546a137058b9.parquet b/questions/MassSlide/dataframes/421875a8a7f441c4bffd546a137058b9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5e6e589f8eca858ec75c08aa28639bc71217d41c --- /dev/null +++ b/questions/MassSlide/dataframes/421875a8a7f441c4bffd546a137058b9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:241df2892aa62d7983f0ffafaa6d2ac340f2ad2fac171429c2b5f0fd9a99d75e +size 5732 diff --git a/questions/MassSlide/dataframes/450660a3c2d24f5bae3eeec80638199c.parquet b/questions/MassSlide/dataframes/450660a3c2d24f5bae3eeec80638199c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cb18e878a03bd8e9b7091bd032657b2fa0133f04 --- /dev/null +++ b/questions/MassSlide/dataframes/450660a3c2d24f5bae3eeec80638199c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcccd2dc32c28a3d219be3486391e3c6ad515acd2f231e1048a42b297d50b0b4 +size 5781 diff --git a/questions/MassSlide/dataframes/477fa4f8ca014f559465bdc2c2318c78.parquet b/questions/MassSlide/dataframes/477fa4f8ca014f559465bdc2c2318c78.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2ad2c00737ef344e5a79ce59f0cd5836f0b87ef9 --- /dev/null +++ b/questions/MassSlide/dataframes/477fa4f8ca014f559465bdc2c2318c78.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8141096d0c2c9eb14f8377d69c1e4c0163d29fb75ad880a0e813125c391cc09 +size 7974 diff --git a/questions/MassSlide/dataframes/4936a5b652995d1cbb5987f3c7ee6866.parquet b/questions/MassSlide/dataframes/4936a5b652995d1cbb5987f3c7ee6866.parquet new file mode 100644 index 0000000000000000000000000000000000000000..35827a178874b33edf3d6ffca8dfd4296af8f681 --- /dev/null +++ b/questions/MassSlide/dataframes/4936a5b652995d1cbb5987f3c7ee6866.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97406a16ac695b6974c2be543adb5d22e442c93114f0dcf4b79a08e78d4cb45d +size 6148 diff --git a/questions/MassSlide/dataframes/4b91015695f94331ba68ab2f9bc8feb1.parquet b/questions/MassSlide/dataframes/4b91015695f94331ba68ab2f9bc8feb1.parquet new file mode 100644 index 0000000000000000000000000000000000000000..20b8dcac6223f5d29c9058c6437c8236a3b95747 --- /dev/null +++ b/questions/MassSlide/dataframes/4b91015695f94331ba68ab2f9bc8feb1.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ebd983cbba38e7508af07af5fb0f0a86bca097f3976c4d756353463d8bbe929c +size 6032 diff --git a/questions/MassSlide/dataframes/509ca665d1854fa596c901a5468197ab.parquet b/questions/MassSlide/dataframes/509ca665d1854fa596c901a5468197ab.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6c334b477113e9fd375dc51c98fb512975d4038b --- /dev/null +++ b/questions/MassSlide/dataframes/509ca665d1854fa596c901a5468197ab.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:204a5f0d03f32c42359275c4604169fa03871fef43df17dcc1fec47c792aba3e +size 5674 diff --git a/questions/MassSlide/dataframes/52cb93e7e69d4abf85d6db293440f76c.parquet b/questions/MassSlide/dataframes/52cb93e7e69d4abf85d6db293440f76c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7557b12d23dae4eda81858e8f3f3573afcbe41e7 --- /dev/null +++ b/questions/MassSlide/dataframes/52cb93e7e69d4abf85d6db293440f76c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:adbb5ae8c670fc865e492e89127330813e26b109328d549e6a065c06ce47d47b +size 5545 diff --git a/questions/MassSlide/dataframes/52d91d5389a652baa478069ea293ca73.parquet b/questions/MassSlide/dataframes/52d91d5389a652baa478069ea293ca73.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7658af32a60186aa85e2123013ce3bb25e04ad02 --- /dev/null +++ b/questions/MassSlide/dataframes/52d91d5389a652baa478069ea293ca73.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8751f01d4abf4032c212c7cbc8930805506a3118a518d35f05acfc2ed139f8a2 +size 7631 diff --git a/questions/MassSlide/dataframes/53d61728255b43158f043258bf085cb9.parquet b/questions/MassSlide/dataframes/53d61728255b43158f043258bf085cb9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a193c1529987cf80978ab8586c8ea4f378316b15 --- /dev/null +++ b/questions/MassSlide/dataframes/53d61728255b43158f043258bf085cb9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:40efa4132686a6955b8417c311be8c113c5b2923a7b14d6f2e4899b277ad5ebc +size 6735 diff --git a/questions/MassSlide/dataframes/59b2dd9b949c42e7ab404612b867f051.parquet b/questions/MassSlide/dataframes/59b2dd9b949c42e7ab404612b867f051.parquet new file mode 100644 index 0000000000000000000000000000000000000000..23e3f75d98f4a0d1ec7b7af42a8426c3feae318b --- /dev/null +++ b/questions/MassSlide/dataframes/59b2dd9b949c42e7ab404612b867f051.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a11ba24d4bb855ff388541321c366e5b3103ee98ea7e25a777ae3776d602034 +size 5997 diff --git a/questions/MassSlide/dataframes/5c6ece223a1a46f6a1987449154ff138.parquet b/questions/MassSlide/dataframes/5c6ece223a1a46f6a1987449154ff138.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d8584ceb53e1801b05f63a8c529b894c0403faf1 --- /dev/null +++ b/questions/MassSlide/dataframes/5c6ece223a1a46f6a1987449154ff138.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:864832338b252e3dd29ad10515cb40eeaa6a1ef02cc120dbd36913951e3f9124 +size 5883 diff --git a/questions/MassSlide/dataframes/5cff394d93ba45fc950eb525d975c078.parquet b/questions/MassSlide/dataframes/5cff394d93ba45fc950eb525d975c078.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a7b81bf5c2e2f5dc7507df4a5e6fed34faa0c0fa --- /dev/null +++ b/questions/MassSlide/dataframes/5cff394d93ba45fc950eb525d975c078.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9f08c5c3a3df25ee1f815dac30c855eab9d53866b0f8ca1a47010807f0328d2 +size 5881 diff --git a/questions/MassSlide/dataframes/633f8b94b4a64afab5c4183b67987092.parquet b/questions/MassSlide/dataframes/633f8b94b4a64afab5c4183b67987092.parquet new file mode 100644 index 0000000000000000000000000000000000000000..08ba984d78af4a99056a0eb140430ad256edc7f0 --- /dev/null +++ b/questions/MassSlide/dataframes/633f8b94b4a64afab5c4183b67987092.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e68a0665112d67552101bdcc319fd59fad019b6afb92b40224a99a747cb5eb00 +size 7660 diff --git a/questions/MassSlide/dataframes/660fdef92d435e0a88bb41b9e958b990.parquet b/questions/MassSlide/dataframes/660fdef92d435e0a88bb41b9e958b990.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0a41a025545f18c84692b00f2322155f2a4ec721 --- /dev/null +++ b/questions/MassSlide/dataframes/660fdef92d435e0a88bb41b9e958b990.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1201cd36b39aad663e8dc0c2588e6edd8ec08cbf09ddd5f0d31dbd1214f5a4fc +size 5521 diff --git a/questions/MassSlide/dataframes/6ff497703dc54edb8534f212c5b7c567.parquet b/questions/MassSlide/dataframes/6ff497703dc54edb8534f212c5b7c567.parquet new file mode 100644 index 0000000000000000000000000000000000000000..392036158fa01f59b8b00e4952382285cdd6c90c --- /dev/null +++ b/questions/MassSlide/dataframes/6ff497703dc54edb8534f212c5b7c567.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0324d062f4741744147f30e009ebf28f8bac8b86f0b326585c63e24e0210219b +size 6106 diff --git a/questions/MassSlide/dataframes/7b689cd67c2d4021930e364b1c2ed8ca.parquet b/questions/MassSlide/dataframes/7b689cd67c2d4021930e364b1c2ed8ca.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d87923b78abe1742f78ca11c0d473a9f68e32363 --- /dev/null +++ b/questions/MassSlide/dataframes/7b689cd67c2d4021930e364b1c2ed8ca.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6cc0dd27dabceb085404427efe4c0e780febcd3cb12e4e39efd53c6912c23caa +size 7974 diff --git a/questions/MassSlide/dataframes/7d1520b2be3c531ab4acbe645c9e18cd.parquet b/questions/MassSlide/dataframes/7d1520b2be3c531ab4acbe645c9e18cd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..02ea87289e45f145916c829de6418b0c5d530e61 --- /dev/null +++ b/questions/MassSlide/dataframes/7d1520b2be3c531ab4acbe645c9e18cd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a3e8371f49b861a44fa34274a2c55e47b96472a075662b564df37e8beba417c3 +size 7967 diff --git a/questions/MassSlide/dataframes/825bb775cf154fdb983653ce20d329dc.parquet b/questions/MassSlide/dataframes/825bb775cf154fdb983653ce20d329dc.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a146a61c8b6eae7f9f2d7e70902b51511c8c4739 --- /dev/null +++ b/questions/MassSlide/dataframes/825bb775cf154fdb983653ce20d329dc.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:566506b4ea0817e6260c6e9c912fd9333c12329fbbadb57c25985b344202ec10 +size 5913 diff --git a/questions/MassSlide/dataframes/833168b23c474c3f95c1ba4d9fff6176.parquet b/questions/MassSlide/dataframes/833168b23c474c3f95c1ba4d9fff6176.parquet new file mode 100644 index 0000000000000000000000000000000000000000..49f065982e88f342cf4f833f50c7fc35f7ffbd90 --- /dev/null +++ b/questions/MassSlide/dataframes/833168b23c474c3f95c1ba4d9fff6176.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58d32176d31166f1e3e3fb8828f32957a2539dfe736a733956cf7bde3d7d0a7a +size 5922 diff --git a/questions/MassSlide/dataframes/8840791a44ca45f2b1e72ee59093cef4.parquet b/questions/MassSlide/dataframes/8840791a44ca45f2b1e72ee59093cef4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..139e9482918fc98844e98e0a64a888e81cb8c94e --- /dev/null +++ b/questions/MassSlide/dataframes/8840791a44ca45f2b1e72ee59093cef4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:315cecb51ccef021059d9c3fe3fe5c122e6750932f4cd1ff4e5a24aa6de6035c +size 6285 diff --git a/questions/MassSlide/dataframes/8aa929de582d5cbf83aa53fe9ab0a174.parquet b/questions/MassSlide/dataframes/8aa929de582d5cbf83aa53fe9ab0a174.parquet new file mode 100644 index 0000000000000000000000000000000000000000..536c37cdce93727f8ec20ce34dc651cca624c9d3 --- /dev/null +++ b/questions/MassSlide/dataframes/8aa929de582d5cbf83aa53fe9ab0a174.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d7fb7d0bffb34dd388a4bb2c13918bb5ce87c707900a747b288a6d221e1d5b47 +size 5672 diff --git a/questions/MassSlide/dataframes/8ffce32ebf95402db33ab77ea1fe5f12.parquet b/questions/MassSlide/dataframes/8ffce32ebf95402db33ab77ea1fe5f12.parquet new file mode 100644 index 0000000000000000000000000000000000000000..efb8b0101da16e6d955a93422fbe4d09920e2747 --- /dev/null +++ b/questions/MassSlide/dataframes/8ffce32ebf95402db33ab77ea1fe5f12.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c9661e0c40d3b7d5c3519f4802ab0661dd80d6655bf48b74f442693c7087d06 +size 5731 diff --git a/questions/MassSlide/dataframes/906b2f75d36d5647a8fe4ef05f19fa5d.parquet b/questions/MassSlide/dataframes/906b2f75d36d5647a8fe4ef05f19fa5d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..86a6084fcbbe5cf178c580dba57282f51f17f030 --- /dev/null +++ b/questions/MassSlide/dataframes/906b2f75d36d5647a8fe4ef05f19fa5d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f4452498b3599ffe38f5ceccf58da092a36065bbf09b2e6f1d16b6a411414d4 +size 5885 diff --git a/questions/MassSlide/dataframes/924138e0d0f9479f8dccc89cc12e589c.parquet b/questions/MassSlide/dataframes/924138e0d0f9479f8dccc89cc12e589c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f7d5ce28e83e0a5c77b8cc0fc81d7134d9f50813 --- /dev/null +++ b/questions/MassSlide/dataframes/924138e0d0f9479f8dccc89cc12e589c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77429e17f94258c0028bc33d520cfdd3d8ae92813102548e321d2bfad94e5328 +size 5675 diff --git a/questions/MassSlide/dataframes/95357a192b404d9d942b699a96f3f91d.parquet b/questions/MassSlide/dataframes/95357a192b404d9d942b699a96f3f91d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cb3748ab341deafbc6560e73887e89568170a5ae --- /dev/null +++ b/questions/MassSlide/dataframes/95357a192b404d9d942b699a96f3f91d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b1642ebaafac2690262f08732e5e1e3257741b557adde5e2d576509c9d0281d +size 7981 diff --git a/questions/MassSlide/dataframes/9646ae5511015044b647887e96eb8dca.parquet b/questions/MassSlide/dataframes/9646ae5511015044b647887e96eb8dca.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3e31ea5432a86d19b9548c7f0d963f7111c33ca9 --- /dev/null +++ b/questions/MassSlide/dataframes/9646ae5511015044b647887e96eb8dca.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e424775177a6cbb6fe2e1c0e0fca783dc6204605b13bee343bbc53ae7af2be1f +size 7981 diff --git a/questions/MassSlide/dataframes/96814fa562ac4bac8cdd1ace014be742.parquet b/questions/MassSlide/dataframes/96814fa562ac4bac8cdd1ace014be742.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c5afef8bcff85e985d815bf7640170d8f62a6254 --- /dev/null +++ b/questions/MassSlide/dataframes/96814fa562ac4bac8cdd1ace014be742.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a120b2f4683f0e225ee2e325194ac06392d68f768e6936bbd8f11d5d21a3bc6 +size 7993 diff --git a/questions/MassSlide/dataframes/97c922e470a6420aaa0db38a7b47ab14.parquet b/questions/MassSlide/dataframes/97c922e470a6420aaa0db38a7b47ab14.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c43cdbed734a047be46078fad296d9b9df08e23e --- /dev/null +++ b/questions/MassSlide/dataframes/97c922e470a6420aaa0db38a7b47ab14.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c73dc37e592a2a88d14e14202158e1c0066058d4caeb2aafe445810656c2fbe5 +size 7686 diff --git a/questions/MassSlide/dataframes/97f285a3809940f68129ed3a9d1d9b26.parquet b/questions/MassSlide/dataframes/97f285a3809940f68129ed3a9d1d9b26.parquet new file mode 100644 index 0000000000000000000000000000000000000000..30a853fbdfee5f8c5719d75f6d2d8d303c3596f5 --- /dev/null +++ b/questions/MassSlide/dataframes/97f285a3809940f68129ed3a9d1d9b26.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:823f86ec3b083a1b40d101d164b36feb6d7a6e0adbfb35f19ce20fb71b5de96e +size 6448 diff --git a/questions/MassSlide/dataframes/9829e3384fae4a77ae3bc21c6c8a6aea.parquet b/questions/MassSlide/dataframes/9829e3384fae4a77ae3bc21c6c8a6aea.parquet new file mode 100644 index 0000000000000000000000000000000000000000..515802cfb04ed4e6ecc002bfa40bd3093c09e700 --- /dev/null +++ b/questions/MassSlide/dataframes/9829e3384fae4a77ae3bc21c6c8a6aea.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dbc6e57c0815825d8f33456f44903e6db9bdb4723cdfdaa10150d97af43a5607 +size 5973 diff --git a/questions/MassSlide/dataframes/9b1982d86e3b5bb892c7d1270f5e8e4c.parquet b/questions/MassSlide/dataframes/9b1982d86e3b5bb892c7d1270f5e8e4c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7c52f8707d64db7ae9a6925b06d4932fde99e827 --- /dev/null +++ b/questions/MassSlide/dataframes/9b1982d86e3b5bb892c7d1270f5e8e4c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12f21cf54689875af9607464f320068c6bf96f66fbd81e1b7d2a159554c9b2e1 +size 5979 diff --git a/questions/MassSlide/dataframes/9b572270b12942a390d67ff9595a2cbd.parquet b/questions/MassSlide/dataframes/9b572270b12942a390d67ff9595a2cbd.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0cfae2218b82dfd89c9d5687836721dc11d3fec1 --- /dev/null +++ b/questions/MassSlide/dataframes/9b572270b12942a390d67ff9595a2cbd.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb18c6fbf7f9caaa86acda8403132286bb50087fc4569f84a66717f5d6b56dd6 +size 8006 diff --git a/questions/MassSlide/dataframes/9e10b79464c24f36a53f252be6541d5f.parquet b/questions/MassSlide/dataframes/9e10b79464c24f36a53f252be6541d5f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..55ababd716fb70ec1744ded70a12046c70c21b1e --- /dev/null +++ b/questions/MassSlide/dataframes/9e10b79464c24f36a53f252be6541d5f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7bd0bdfa2f0d7dc187a109116586f7f6dbade5e6b0489674d52c625906c4c100 +size 7386 diff --git a/questions/MassSlide/dataframes/9e2271b4625541f7964a03844c658511.parquet b/questions/MassSlide/dataframes/9e2271b4625541f7964a03844c658511.parquet new file mode 100644 index 0000000000000000000000000000000000000000..36342ab1be02d3d5975dda9eb93b6fac4defd939 --- /dev/null +++ b/questions/MassSlide/dataframes/9e2271b4625541f7964a03844c658511.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42e00fba61b3bd39c8d85f79eab8f4975d8fa9d9f8ec75f2da643384591fc955 +size 8014 diff --git a/questions/MassSlide/dataframes/a3dc646b3f7647079bfb49c9acaa18e6.parquet b/questions/MassSlide/dataframes/a3dc646b3f7647079bfb49c9acaa18e6.parquet new file mode 100644 index 0000000000000000000000000000000000000000..70584b2ddea813de684da64b0bcd0538d06562a5 --- /dev/null +++ b/questions/MassSlide/dataframes/a3dc646b3f7647079bfb49c9acaa18e6.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b4acddf846836cae5f030d8cce1035d08d82f46d434d4d78f5928802519646c +size 5814 diff --git a/questions/MassSlide/dataframes/a788bcbddb3e4143afc049ae6fd613ed.parquet b/questions/MassSlide/dataframes/a788bcbddb3e4143afc049ae6fd613ed.parquet new file mode 100644 index 0000000000000000000000000000000000000000..db40eb7614f067f8d7e49df40108b613bee0a18e --- /dev/null +++ b/questions/MassSlide/dataframes/a788bcbddb3e4143afc049ae6fd613ed.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6cf3b10968331a43d104b06ba0c0d6c7375ba4ed11d375e1e58fd0091cd71f1d +size 7974 diff --git a/questions/MassSlide/dataframes/ab1dbdc0c2344be7a4b105dd445c3e62.parquet b/questions/MassSlide/dataframes/ab1dbdc0c2344be7a4b105dd445c3e62.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7c1a53356a76c54d6c4c51eb3aea222721a204b1 --- /dev/null +++ b/questions/MassSlide/dataframes/ab1dbdc0c2344be7a4b105dd445c3e62.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4007bdd488705de440e5c78cb53f5627b87f5de45226664d815d2399f3d43a7f +size 6095 diff --git a/questions/MassSlide/dataframes/aec60d90f7ef4979a80c56f5c5bdeafa.parquet b/questions/MassSlide/dataframes/aec60d90f7ef4979a80c56f5c5bdeafa.parquet new file mode 100644 index 0000000000000000000000000000000000000000..7a3b0552509a6c47548b0b8e073ac7f341250319 --- /dev/null +++ b/questions/MassSlide/dataframes/aec60d90f7ef4979a80c56f5c5bdeafa.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d2f3c981a37e35d40852c89d305bac35d6fba110d1b90e40b673cb46c5f4f6e +size 6564 diff --git a/questions/MassSlide/dataframes/afec7fdef34d4256a0060406c9b1cd7c.parquet b/questions/MassSlide/dataframes/afec7fdef34d4256a0060406c9b1cd7c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ed805bea4a456af8222f503883dd537e11e3d65a --- /dev/null +++ b/questions/MassSlide/dataframes/afec7fdef34d4256a0060406c9b1cd7c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e752f045af92242e30e9dbd9002efce7fe98a4e510565fdd08d8f923e0f14fe +size 6629 diff --git a/questions/MassSlide/dataframes/b1fb837092144b5cae1c2a952a2ea10c.parquet b/questions/MassSlide/dataframes/b1fb837092144b5cae1c2a952a2ea10c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9bb9111c6429ca60de38781a2b679e60e960f8fb --- /dev/null +++ b/questions/MassSlide/dataframes/b1fb837092144b5cae1c2a952a2ea10c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9d2f412b240a15b5ddf946c5f0ddbc1a66b9e2d381f17708b8cc5dbb7d0b590 +size 6100 diff --git a/questions/MassSlide/dataframes/b6a00423859b48e8b634aaf7a4b4208a.parquet b/questions/MassSlide/dataframes/b6a00423859b48e8b634aaf7a4b4208a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f7b05b2210478bd861d27aad0bce27cce87fcfa6 --- /dev/null +++ b/questions/MassSlide/dataframes/b6a00423859b48e8b634aaf7a4b4208a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47c3024e5652b859ca62f194c501d9e6da0940d34ce6dd6ad226fd8d70766157 +size 6331 diff --git a/questions/MassSlide/dataframes/b6e5860643684682b60e6c146121b2be.parquet b/questions/MassSlide/dataframes/b6e5860643684682b60e6c146121b2be.parquet new file mode 100644 index 0000000000000000000000000000000000000000..d15f0dd23cc2d0bb1060b2acd8793cc3b5ae2980 --- /dev/null +++ b/questions/MassSlide/dataframes/b6e5860643684682b60e6c146121b2be.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96d9a54abab1c343dee0b157ae2c8c6299270b2bb0b0334d900c0c4f72106d0e +size 5843 diff --git a/questions/MassSlide/dataframes/b794f7c284485e548f68445b0c123a3a.parquet b/questions/MassSlide/dataframes/b794f7c284485e548f68445b0c123a3a.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c3612ddbb11d4e5fd6d80fcbb713a0d69807c46f --- /dev/null +++ b/questions/MassSlide/dataframes/b794f7c284485e548f68445b0c123a3a.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66514e2889d810dcb5d3f7bdeaef23af56bcf1fb0ef767aacf0b2fb1c3e3e8ff +size 5619 diff --git a/questions/MassSlide/dataframes/b7d2dbe41aef52c58ccf2a7fd13903a5.parquet b/questions/MassSlide/dataframes/b7d2dbe41aef52c58ccf2a7fd13903a5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3663494e4a5dd055ccc68d315914633b42264da5 --- /dev/null +++ b/questions/MassSlide/dataframes/b7d2dbe41aef52c58ccf2a7fd13903a5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42419234864b7115574d9ae4caaf790bc0f6f7e08b7cc75051045c1cef429416 +size 7981 diff --git a/questions/MassSlide/dataframes/b9d5d04e374c47d58d820fca9d364a11.parquet b/questions/MassSlide/dataframes/b9d5d04e374c47d58d820fca9d364a11.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e39515828ed2c8868dd5838eab52d5cf6128c767 --- /dev/null +++ b/questions/MassSlide/dataframes/b9d5d04e374c47d58d820fca9d364a11.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4bfa28a58276f22e807f085cdae2a235261569b3da8ae129009450c23d771d56 +size 6729 diff --git a/questions/MassSlide/dataframes/b9f73c1ba2e14456bc1cb2701b0d61d4.parquet b/questions/MassSlide/dataframes/b9f73c1ba2e14456bc1cb2701b0d61d4.parquet new file mode 100644 index 0000000000000000000000000000000000000000..f389251356e995026cc8cbb2eb61b20c66694b61 --- /dev/null +++ b/questions/MassSlide/dataframes/b9f73c1ba2e14456bc1cb2701b0d61d4.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ed5d95465da77ab7d89564a5786f89b09ae74728938f4dc899170fc417cc76f6 +size 5570 diff --git a/questions/MassSlide/dataframes/badd89b9580a4d319ec3f8a0bee814b1.parquet b/questions/MassSlide/dataframes/badd89b9580a4d319ec3f8a0bee814b1.parquet new file mode 100644 index 0000000000000000000000000000000000000000..81e3a2ce245d54ffa4b6c256953469a4c863ecd1 --- /dev/null +++ b/questions/MassSlide/dataframes/badd89b9580a4d319ec3f8a0bee814b1.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb9b1735d5a6fc32ac55d944196546de3a35eabff7ca8b6a87c9f2edf1b3145a +size 7999 diff --git a/questions/MassSlide/dataframes/baf1e1790569451993d611a8cb903120.parquet b/questions/MassSlide/dataframes/baf1e1790569451993d611a8cb903120.parquet new file mode 100644 index 0000000000000000000000000000000000000000..682efa32bae18754f5650bbf3fc28d1685e43bec --- /dev/null +++ b/questions/MassSlide/dataframes/baf1e1790569451993d611a8cb903120.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e296c0e04c4a880a353de3414ff91c5e1a5deddff249c4f9655abdcf0d98c15 +size 7981 diff --git a/questions/MassSlide/dataframes/bb5c824b74c65fd0bfdeeb1b5969e46c.parquet b/questions/MassSlide/dataframes/bb5c824b74c65fd0bfdeeb1b5969e46c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..dae0163057144af9d0fc32b760194101524d7684 --- /dev/null +++ b/questions/MassSlide/dataframes/bb5c824b74c65fd0bfdeeb1b5969e46c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62d3e1802e5b119769ce61aff7c7a906430b5b649e9d48dbe522f9e8489dbb1e +size 7987 diff --git a/questions/MassSlide/dataframes/bd084d19f2444af8a6f98ec7815d2882.parquet b/questions/MassSlide/dataframes/bd084d19f2444af8a6f98ec7815d2882.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6dea238d998edb2869ca083b1da1fc5d81b673c9 --- /dev/null +++ b/questions/MassSlide/dataframes/bd084d19f2444af8a6f98ec7815d2882.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:641e8470db797c0013dc393e88a067fff02b07b655db6ba9d203d9fd6d054033 +size 6783 diff --git a/questions/MassSlide/dataframes/bd1db72c3cdb481caa778ac3639cccb3.parquet b/questions/MassSlide/dataframes/bd1db72c3cdb481caa778ac3639cccb3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..482bec387be2208ca2b25577010e4c8a064e4ef6 --- /dev/null +++ b/questions/MassSlide/dataframes/bd1db72c3cdb481caa778ac3639cccb3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7415788836c1a558703517523361304f1464f521c5d929601d11de3b531e7c99 +size 5981 diff --git a/questions/MassSlide/dataframes/bd222ab0a4034903923ada8b7a09816f.parquet b/questions/MassSlide/dataframes/bd222ab0a4034903923ada8b7a09816f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..cf1065e619c8c8fe34d1e1a69f65d6b256750423 --- /dev/null +++ b/questions/MassSlide/dataframes/bd222ab0a4034903923ada8b7a09816f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:236d6b5b3ef1fa17594c894369c41185786b1fdaba33688491be0a4c1fa79dfa +size 6210 diff --git a/questions/MassSlide/dataframes/bdf601681b8541659ad53f2b52d11526.parquet b/questions/MassSlide/dataframes/bdf601681b8541659ad53f2b52d11526.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9f5eab05b433a8c1634cd90f0e2e36f6d2459094 --- /dev/null +++ b/questions/MassSlide/dataframes/bdf601681b8541659ad53f2b52d11526.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5d114284594339afe20ac960b4c568aff4391b16688e0a361420273c7754cde +size 6078 diff --git a/questions/MassSlide/dataframes/c1e94ce33230421a847d36495497409d.parquet b/questions/MassSlide/dataframes/c1e94ce33230421a847d36495497409d.parquet new file mode 100644 index 0000000000000000000000000000000000000000..898d800962db47a153e85a4b0d040fb4e77d0d5b --- /dev/null +++ b/questions/MassSlide/dataframes/c1e94ce33230421a847d36495497409d.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d7cdf0704dfa8a82d08c8278870c0414a577a655d7f6c9057ee24fc5f7544126 +size 7967 diff --git a/questions/MassSlide/dataframes/c56def5d135e40c290812174a4d3a066.parquet b/questions/MassSlide/dataframes/c56def5d135e40c290812174a4d3a066.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3b6745f783d48f4d62c6069e13251558a693c1c9 --- /dev/null +++ b/questions/MassSlide/dataframes/c56def5d135e40c290812174a4d3a066.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f460bddfa110a3dcb2af5344b94b3cc5fad917bc643e0c5c433df9bc676fd64 +size 7967 diff --git a/questions/MassSlide/dataframes/c5d7959f17da52dda612bcc328e0b427.parquet b/questions/MassSlide/dataframes/c5d7959f17da52dda612bcc328e0b427.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b1ad2e5d37375249a0b824f8953ea4298a8ca7bc --- /dev/null +++ b/questions/MassSlide/dataframes/c5d7959f17da52dda612bcc328e0b427.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06c5a3e7e0c6cb87fc1f94ffa465a2dcc8c8a38a77c185946482b3d08a4fd829 +size 7981 diff --git a/questions/MassSlide/dataframes/c6cc5bbcb54c5d37b6f62836386f8f14.parquet b/questions/MassSlide/dataframes/c6cc5bbcb54c5d37b6f62836386f8f14.parquet new file mode 100644 index 0000000000000000000000000000000000000000..125492150ce2cc217cd0e3e9c1c7740cb52d28e0 --- /dev/null +++ b/questions/MassSlide/dataframes/c6cc5bbcb54c5d37b6f62836386f8f14.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae7bfc2582cc2fc385222b6ba98da4f082524f2fb0a0f1c5a784f0a399df8f76 +size 7981 diff --git a/questions/MassSlide/dataframes/ca2826a9ea444e4fbc8b2cd785dfc4a3.parquet b/questions/MassSlide/dataframes/ca2826a9ea444e4fbc8b2cd785dfc4a3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..9969e84f77b17fd4c35a868c78870e02e0baa727 --- /dev/null +++ b/questions/MassSlide/dataframes/ca2826a9ea444e4fbc8b2cd785dfc4a3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6dcf3aa852d54a6196363d5802b07bc30350b3e749354dc3113e02ed0271fd12 +size 7974 diff --git a/questions/MassSlide/dataframes/cc910fa7368c5587804f005866a22079.parquet b/questions/MassSlide/dataframes/cc910fa7368c5587804f005866a22079.parquet new file mode 100644 index 0000000000000000000000000000000000000000..2e4d661558c2dff95cb8c3380d44ffcd8faa57c7 --- /dev/null +++ b/questions/MassSlide/dataframes/cc910fa7368c5587804f005866a22079.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1ebc8b3c98c5511aa05486397f906392f8319ce3e7543a13d78c4fae2decedd +size 5620 diff --git a/questions/MassSlide/dataframes/cd5b8bdef4d44b7d8f6682e3cde5fcf2.parquet b/questions/MassSlide/dataframes/cd5b8bdef4d44b7d8f6682e3cde5fcf2.parquet new file mode 100644 index 0000000000000000000000000000000000000000..19558b3e3d0bb65c91c58006d57c3600efc50603 --- /dev/null +++ b/questions/MassSlide/dataframes/cd5b8bdef4d44b7d8f6682e3cde5fcf2.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:527bcb6a50ae7545d9ceec272640c2823f473467c0e5c09589138ac49a2b42e6 +size 7981 diff --git a/questions/MassSlide/dataframes/cfcf113e827240a6b7e7a4be86078864.parquet b/questions/MassSlide/dataframes/cfcf113e827240a6b7e7a4be86078864.parquet new file mode 100644 index 0000000000000000000000000000000000000000..897d60fb3ca0b94d967033d3f59f6e0e4d7ccaf3 --- /dev/null +++ b/questions/MassSlide/dataframes/cfcf113e827240a6b7e7a4be86078864.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbb9330aa9532da0911849072d810fb08f88974836df82444e3a4d09b3454fac +size 7981 diff --git a/questions/MassSlide/dataframes/d1141599d62d4cd8a57708873e904583.parquet b/questions/MassSlide/dataframes/d1141599d62d4cd8a57708873e904583.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4e6eeda33fd1e7b63a39bd93cb3b0b988dfb6565 --- /dev/null +++ b/questions/MassSlide/dataframes/d1141599d62d4cd8a57708873e904583.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f55342d8f69011b166325d2d7e051cb11f39ee413c6d44c0a7c621eb1276d80 +size 7910 diff --git a/questions/MassSlide/dataframes/d2354df65fff5ddf9e1b8e5d138282bc.parquet b/questions/MassSlide/dataframes/d2354df65fff5ddf9e1b8e5d138282bc.parquet new file mode 100644 index 0000000000000000000000000000000000000000..6cece67b5ce3d06aab32e99b29ce470ae52d0990 --- /dev/null +++ b/questions/MassSlide/dataframes/d2354df65fff5ddf9e1b8e5d138282bc.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:328d8c945d70bf30708af1e71eac4c4c76584620f46fe6ef2b985556ba6c44ea +size 7987 diff --git a/questions/MassSlide/dataframes/d8108f5dafb05591a0165983da77f78c.parquet b/questions/MassSlide/dataframes/d8108f5dafb05591a0165983da77f78c.parquet new file mode 100644 index 0000000000000000000000000000000000000000..ec45899e727ba0a7124a389440f234f3b372baff --- /dev/null +++ b/questions/MassSlide/dataframes/d8108f5dafb05591a0165983da77f78c.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22fdfd34285b0733026852bb0104eac509421e83c931d4d234f0f657da22ff41 +size 5523 diff --git a/questions/MassSlide/dataframes/d913f421c0964aea99408cec015c64c1.parquet b/questions/MassSlide/dataframes/d913f421c0964aea99408cec015c64c1.parquet new file mode 100644 index 0000000000000000000000000000000000000000..79f89dfc283329ef911e698d4e9a90e70062580f --- /dev/null +++ b/questions/MassSlide/dataframes/d913f421c0964aea99408cec015c64c1.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7977a2e99588a96a1a4c377214bebc0c89d38008b1cf7cbd06e7957e08233b79 +size 5690 diff --git a/questions/MassSlide/dataframes/d950be4ceaf855e4b99175c9b68e37f5.parquet b/questions/MassSlide/dataframes/d950be4ceaf855e4b99175c9b68e37f5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e578c66f7168e5ff47d3b90b360efa3343b47a7b --- /dev/null +++ b/questions/MassSlide/dataframes/d950be4ceaf855e4b99175c9b68e37f5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e3cae7e6381afb21ffca5c140725b9eb493dcea27322874ba58c89a511400ef +size 7981 diff --git a/questions/MassSlide/dataframes/dcc54ad4f3014c0cb2545e1f84a9ee38.parquet b/questions/MassSlide/dataframes/dcc54ad4f3014c0cb2545e1f84a9ee38.parquet new file mode 100644 index 0000000000000000000000000000000000000000..bf272dc27ce63f958d1335476bc621c88120b053 --- /dev/null +++ b/questions/MassSlide/dataframes/dcc54ad4f3014c0cb2545e1f84a9ee38.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c403ffbbbf713b39083c765da12ba87294cdc22dd5ecb956bb43f1f1e875303 +size 7512 diff --git a/questions/MassSlide/dataframes/dd028d3ab7fb4aae84694d21ecf4c201.parquet b/questions/MassSlide/dataframes/dd028d3ab7fb4aae84694d21ecf4c201.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5dcd8592585afc67b676f8fe14a624763e3b269f --- /dev/null +++ b/questions/MassSlide/dataframes/dd028d3ab7fb4aae84694d21ecf4c201.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6744a19fe37451396ffa5ab08b539fd89063138dede6b5c00cab04c44863d3ef +size 7978 diff --git a/questions/MassSlide/dataframes/e07a5edf78e844b8ba42f7b4add2cfb7.parquet b/questions/MassSlide/dataframes/e07a5edf78e844b8ba42f7b4add2cfb7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..60b96a1a72714357fcec4f2dd854387deb124f0c --- /dev/null +++ b/questions/MassSlide/dataframes/e07a5edf78e844b8ba42f7b4add2cfb7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb6ffe0dd49c21d84572a3a18898c97f75ecd2caa2e5da8f568575d3179e331a +size 6388 diff --git a/questions/MassSlide/dataframes/e0934b52a37759e3a90cd51c68f250e5.parquet b/questions/MassSlide/dataframes/e0934b52a37759e3a90cd51c68f250e5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..a140e78758bb984b0a2e3e51f4b3776e105d559d --- /dev/null +++ b/questions/MassSlide/dataframes/e0934b52a37759e3a90cd51c68f250e5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d51b119722505ef8c20725486ecddb3edf70e4eb842592e176f0ef78dc675eca +size 7981 diff --git a/questions/MassSlide/dataframes/e2cc7f4083515d948057472541e749f9.parquet b/questions/MassSlide/dataframes/e2cc7f4083515d948057472541e749f9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..5b7e2296f3d8eedc1c20ad1dbe01e253b13640b7 --- /dev/null +++ b/questions/MassSlide/dataframes/e2cc7f4083515d948057472541e749f9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d6411913cfb609c8935d7e3901cf1efa6f411d3d18262a17a531b5565ebf451 +size 7980 diff --git a/questions/MassSlide/dataframes/e40e7855b79058599803ffe83e69df2f.parquet b/questions/MassSlide/dataframes/e40e7855b79058599803ffe83e69df2f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..4d30075fd66c4c6bb4081574e8add9a45b7d73e8 --- /dev/null +++ b/questions/MassSlide/dataframes/e40e7855b79058599803ffe83e69df2f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8dc77d3525b2b046ae1ea63993217ae29d86d01cd1dea63c9d514c0fde63b241 +size 7974 diff --git a/questions/MassSlide/dataframes/e4156c6755bc5b86ab4fb7c158deb515.parquet b/questions/MassSlide/dataframes/e4156c6755bc5b86ab4fb7c158deb515.parquet new file mode 100644 index 0000000000000000000000000000000000000000..57934da1105270ddc0ca3393354edcc459c068c9 --- /dev/null +++ b/questions/MassSlide/dataframes/e4156c6755bc5b86ab4fb7c158deb515.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5ac958b48524a15867e35aaa170edd31accd1d4770c52d3e5a7b6d512b09bcc +size 7974 diff --git a/questions/MassSlide/dataframes/e430606893ee431087b189d71cd29ac5.parquet b/questions/MassSlide/dataframes/e430606893ee431087b189d71cd29ac5.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0671a0ac9a316caf05222ae17e5ee5ee5f77709e --- /dev/null +++ b/questions/MassSlide/dataframes/e430606893ee431087b189d71cd29ac5.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3bbe84c0bbcdfb651d6e8ba9a6ff0f297d370404874db84ae589cf3ac30a1b0 +size 7139 diff --git a/questions/MassSlide/dataframes/e459f3032a4152a496af1745a8b75c10.parquet b/questions/MassSlide/dataframes/e459f3032a4152a496af1745a8b75c10.parquet new file mode 100644 index 0000000000000000000000000000000000000000..fe5f58bc5a8dc357ab57b4540012260bb30d46d4 --- /dev/null +++ b/questions/MassSlide/dataframes/e459f3032a4152a496af1745a8b75c10.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b1259470389c2f1c0d40a5fa8cdf6a0b76aded6fb8b08756866e110f38fc790 +size 5664 diff --git a/questions/MassSlide/dataframes/e4a4fea8b64e496cb7e24f4de9f270b7.parquet b/questions/MassSlide/dataframes/e4a4fea8b64e496cb7e24f4de9f270b7.parquet new file mode 100644 index 0000000000000000000000000000000000000000..e4b723388eff68dd02cefac24c2a8cab2c9343f1 --- /dev/null +++ b/questions/MassSlide/dataframes/e4a4fea8b64e496cb7e24f4de9f270b7.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb8cce5dbf11f67be903b95a6c0f9345100355510c1c831b5aa2ca61c9e57b42 +size 7936 diff --git a/questions/MassSlide/dataframes/e4faa5a0e1ad4ca4b94bf2f18f4a093f.parquet b/questions/MassSlide/dataframes/e4faa5a0e1ad4ca4b94bf2f18f4a093f.parquet new file mode 100644 index 0000000000000000000000000000000000000000..faee7dd7ddd2254f5cea3de1218767f711549f98 --- /dev/null +++ b/questions/MassSlide/dataframes/e4faa5a0e1ad4ca4b94bf2f18f4a093f.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b318d6166aac73a16e36d20f4616b01a9ed2d0ed900eec2a4295fa1d15a06c6a +size 6398 diff --git a/questions/MassSlide/dataframes/e6e67852460340c291d3b2d33978e955.parquet b/questions/MassSlide/dataframes/e6e67852460340c291d3b2d33978e955.parquet new file mode 100644 index 0000000000000000000000000000000000000000..be0a11f787fc17422e504decfdad66a524194a8c --- /dev/null +++ b/questions/MassSlide/dataframes/e6e67852460340c291d3b2d33978e955.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c11032bdb36349d895631f9b8e9fec54a88d20052363b3e07a508cf548513ab9 +size 6971 diff --git a/questions/MassSlide/dataframes/e803fdb9e638552d9050b896808b9b46.parquet b/questions/MassSlide/dataframes/e803fdb9e638552d9050b896808b9b46.parquet new file mode 100644 index 0000000000000000000000000000000000000000..374165d8745c93717d5e0a914dfb79737bb745cb --- /dev/null +++ b/questions/MassSlide/dataframes/e803fdb9e638552d9050b896808b9b46.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6aefaa32470f1429e5ee8ce056aed15025b49765fa2405aae648b5d0563482b +size 7967 diff --git a/questions/MassSlide/dataframes/e8e9190868644fc38aa1113756e04fee.parquet b/questions/MassSlide/dataframes/e8e9190868644fc38aa1113756e04fee.parquet new file mode 100644 index 0000000000000000000000000000000000000000..62db344a4e6412325f8ff97060fc67188969c107 --- /dev/null +++ b/questions/MassSlide/dataframes/e8e9190868644fc38aa1113756e04fee.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1e02db3767aa8ba3ce31af9923e6ed59b5f0645c4bfe50c826004825db2cf72 +size 7882 diff --git a/questions/MassSlide/dataframes/ead1bd72fabc53758d90a08e1dcd2507.parquet b/questions/MassSlide/dataframes/ead1bd72fabc53758d90a08e1dcd2507.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0c0128ecb60695e688a1a14ce3e0fca5ad5bf5cb --- /dev/null +++ b/questions/MassSlide/dataframes/ead1bd72fabc53758d90a08e1dcd2507.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58695a514ed999add58dae5e828ad3ade1941fdb83d025dff81936e88cf5aba6 +size 7980 diff --git a/questions/MassSlide/dataframes/f2f13ebd94865bf28ae7eeea4ef29cdf.parquet b/questions/MassSlide/dataframes/f2f13ebd94865bf28ae7eeea4ef29cdf.parquet new file mode 100644 index 0000000000000000000000000000000000000000..894d46c9b36046d4dd62c840411cbb236aab58ec --- /dev/null +++ b/questions/MassSlide/dataframes/f2f13ebd94865bf28ae7eeea4ef29cdf.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:758a404ac14121151979a3878bc8a7bbd7725b07864827d1a5d80aa43d6ed6d4 +size 7974 diff --git a/questions/MassSlide/dataframes/f3198e91ff7d593b8b1d00e99b154708.parquet b/questions/MassSlide/dataframes/f3198e91ff7d593b8b1d00e99b154708.parquet new file mode 100644 index 0000000000000000000000000000000000000000..b91d70b339faa23c2d2a1bb712fd51dd1b7b0c3a --- /dev/null +++ b/questions/MassSlide/dataframes/f3198e91ff7d593b8b1d00e99b154708.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6978d0ed0cac90283cb4c2dd9af606b784b519ab98e6f8f2558926af8a0f8be6 +size 7971 diff --git a/questions/MassSlide/dataframes/f664534ad3bb423eae4c699ad1d1ad79.parquet b/questions/MassSlide/dataframes/f664534ad3bb423eae4c699ad1d1ad79.parquet new file mode 100644 index 0000000000000000000000000000000000000000..c59a627e2ee11f78690d4dd6203c69f503b76294 --- /dev/null +++ b/questions/MassSlide/dataframes/f664534ad3bb423eae4c699ad1d1ad79.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:948f096ea3d05bc1af9f1083a47d04217de3fc271a9eff81e256e2969b026138 +size 5688 diff --git a/questions/MassSlide/dataframes/fb2367ca63f6468bbb95f4ddbd0d99c9.parquet b/questions/MassSlide/dataframes/fb2367ca63f6468bbb95f4ddbd0d99c9.parquet new file mode 100644 index 0000000000000000000000000000000000000000..29a706c47544256f6a66f3f9cdb505e34bcb8941 --- /dev/null +++ b/questions/MassSlide/dataframes/fb2367ca63f6468bbb95f4ddbd0d99c9.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76409200d8ce92ff460a92a0472bb1b0003cda09e661a224794f30c8d476b67a +size 6023 diff --git a/questions/MassSlide/dataframes/ff3a19d5eaad46c1aed4ca4373c26de3.parquet b/questions/MassSlide/dataframes/ff3a19d5eaad46c1aed4ca4373c26de3.parquet new file mode 100644 index 0000000000000000000000000000000000000000..3852004aec6bb575a0dda1f6af18311c56032c36 --- /dev/null +++ b/questions/MassSlide/dataframes/ff3a19d5eaad46c1aed4ca4373c26de3.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d49277cfcea9fcd5f11be4d2ed2c0941ab57d7c2cd0fd154b1cebc9b0417d43 +size 7981 diff --git a/questions/MassSlide/dataframes/ffcb3e01741a516091f8da7aeb689274.parquet b/questions/MassSlide/dataframes/ffcb3e01741a516091f8da7aeb689274.parquet new file mode 100644 index 0000000000000000000000000000000000000000..0bfcf48e48405458791e9702cfbf53db6ffadbb6 --- /dev/null +++ b/questions/MassSlide/dataframes/ffcb3e01741a516091f8da7aeb689274.parquet @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76092a319e17fe87b07091b968d05a8f6eb68cc99d8c348462f42c43753c0a1b +size 6170 diff --git a/questions/MassSlide/model_record.json b/questions/MassSlide/model_record.json new file mode 100644 index 0000000000000000000000000000000000000000..be0e9fc81c6baa1a18e738fa9eddb7910acdf4f2 --- /dev/null +++ b/questions/MassSlide/model_record.json @@ -0,0 +1,7508 @@ +{ + "c1e94ce33230421a847d36495497409d": { + "parameters_hash": "700d0f17f4c708766823cd85a5c5a833", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T16:15:13.891102", + "end_time_simulation": 9.979999542236328 + }, + "6d036dc6b94b41a3984fc3d355a3bdf8": { + "parameters_hash": "e31c9d2245fabaeeceab8be874af154e", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T16:15:20.987673", + "end_time_simulation": 9.979999542236328 + }, + "d60501856a604718aa07ba766010b421": { + "parameters_hash": "8bdb49f12fa0cecca925f9b8f27bd1c4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:15:24.590162", + "end_time_simulation": 9.979999542236328 + }, + "218e2eadaa8d45ac943dec78abed5708": { + "parameters_hash": "e2a01c786ba899f6a89c3e280de0a3cc", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T16:15:27.635311", + "end_time_simulation": 9.979999542236328 + }, + "aa8835a677c2524babff44473e38fd8c": { + "parameters_hash": "d5d4ba3ebbb0b44825c1463cdf8c0772", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:15:30.645908", + "end_time_simulation": 9.979999542236328 + }, + "43de5f825be1499f8562921d4aaf0843": { + "parameters_hash": "ccf4b1f01fe6b19388e8fa41c0b02b92", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T16:15:33.453802", + "end_time_simulation": 9.979999542236328 + }, + "60ab04c8f5ae4832ae0a9426d6ccb8d9": { + "parameters_hash": "8435f58e8c054a67bb1e40d6597bba86", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:15:35.952662", + "end_time_simulation": 9.979999542236328 + }, + "63a6981b6cdc4f23a2aef13645a7d99d": { + "parameters_hash": "a0147a07a621af74f4b080acb60ceead", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T16:15:38.808803", + "end_time_simulation": 9.979999542236328 + }, + "a749935a03ea4665adb3fe2b8513aea5": { + "parameters_hash": "9eb410abddf925859ffdaf49c9fa2216", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:15:41.701585", + "end_time_simulation": 9.979999542236328 + }, + "0da3463ffd384d06b46b79818faaf7f7": { + "parameters_hash": "258d63e595149d7aa041fa8f31ba1659", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:15:45.439095", + "end_time_simulation": 9.979999542236328 + }, + "9c313f57b5765949aa47e03e71b8149c": { + "parameters_hash": "b31c36eacf1cd7c9f2d1f6d115fad54d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:15:48.237570", + "end_time_simulation": 9.979999542236328 + }, + "ca2826a9ea444e4fbc8b2cd785dfc4a3": { + "parameters_hash": "67183565fd9cac0a84d61b80d05ca8f1", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:20:03.521044", + "end_time_simulation": 9.979999542236328 + }, + "0bd09a8e806758cfbeacdbc6f2757b0f": { + "parameters_hash": "21093296938fd5dc5d68e73469e9b9a5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:20:06.287044", + "end_time_simulation": 9.979999542236328 + }, + "8ab3995dfe304543930706b96ee89ab8": { + "parameters_hash": "eeaa582a407023225d2bb6dd05baa739", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:15:51.169442", + "end_time_simulation": 9.979999542236328 + }, + "83ca6beeeb7d5cadb1203f91bf5bf488": { + "parameters_hash": "66d7a6ea866d94f82c863a03ab3bdb8b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:15:54.131924", + "end_time_simulation": 9.979999542236328 + }, + "85f884b40f5947e8a72dd8e16f10470a": { + "parameters_hash": "968a31d19679bf5ad2c7dff0992cc874", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:15:56.940694", + "end_time_simulation": 9.979999542236328 + }, + "2cd7455da599587484f523c659acd5ff": { + "parameters_hash": "c49d90c31d845301cbd169af90d43c6c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:15:59.170080", + "end_time_simulation": 9.979999542236328 + }, + "253c13b3eacb4fd2a7778e99a728fd0f": { + "parameters_hash": "288affbc9497ca33c31d4df65b78067b", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:34:58.485865", + "end_time_simulation": 9.979999542236328 + }, + "94f89a68fbcb43238ee398e70350b970": { + "parameters_hash": "c5033fc92f612e9cbf94a4e206867507", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:35:01.825463", + "end_time_simulation": 9.979999542236328 + }, + "fca71840d31c4154bd013a8854372cef": { + "parameters_hash": "e4cfb0271cb447f9aded3fcbdb7d85b9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:04.219619", + "end_time_simulation": 9.979999542236328 + }, + "4b91015695f94331ba68ab2f9bc8feb1": { + "parameters_hash": "c4cdd6d4e155b756f2c56d0daf860e27", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:35:06.980563", + "end_time_simulation": 9.979999542236328 + }, + "c2700a254b2f565c9ac55004c45e4c64": { + "parameters_hash": "a093d942868d25d06bd367fb3c07bc0c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:09.665897", + "end_time_simulation": 9.979999542236328 + }, + "bd2ad578f38442198c43b53f57f7e5bc": { + "parameters_hash": "65191660686c0f70f80cf0acc378809f", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:35:12.560816", + "end_time_simulation": 9.979999542236328 + }, + "4209f1dac91e437a814a09e8c8309701": { + "parameters_hash": "dabc267fd8ae9b6e4daf6b50b23e2b78", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:15.214651", + "end_time_simulation": 9.979999542236328 + }, + "91ecc4b1560844b79b5875386a25823e": { + "parameters_hash": "cee49150f44c5146ccee22425b819184", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:35:18.073845", + "end_time_simulation": 9.979999542236328 + }, + "d378f3c7b0f35f65a78028a5d3f21f2d": { + "parameters_hash": "b583df0ca5b3d11888ea8ea77292cc0f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:20.813055", + "end_time_simulation": 9.979999542236328 + }, + "3a7fe96753954c64b486e313293efecb": { + "parameters_hash": "149216a1d7a811bfcdade6c44b24bc1c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:35:24.245757", + "end_time_simulation": 9.979999542236328 + }, + "66210a7b87ca4e75af7d8ba85d06d167": { + "parameters_hash": "a4270ef1347a98b196f01f58ef8c3481", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:27.464067", + "end_time_simulation": 9.979999542236328 + }, + "9a08550baa114f9b995165274bdc1554": { + "parameters_hash": "2abe6b24920d6d2531c3dd27a000b86a", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:35:31.054151", + "end_time_simulation": 9.979999542236328 + }, + "b2c587e579b45876842fe2da7990f07f": { + "parameters_hash": "b530d09d8e8491b4b93fb254740ed8a0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:33.995689", + "end_time_simulation": 9.979999542236328 + }, + "45071de2497d4cdf80df422c7323745e": { + "parameters_hash": "7234b414e966bd405c2c6fea88fc7bc8", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:35:36.690389", + "end_time_simulation": 9.979999542236328 + }, + "80907ada2e2a4f0a999a8477bfe4617d": { + "parameters_hash": "72b4bddabfeee5b189ace2e324a30051", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:39.335635", + "end_time_simulation": 9.979999542236328 + }, + "2fd479b55fad4157a6498952b8b956d5": { + "parameters_hash": "5b3624bf11add0fdd034a0c8f64f070d", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:35:41.847729", + "end_time_simulation": 9.979999542236328 + }, + "647272b561784ff7a01c3c20ec40e388": { + "parameters_hash": "3a0928c0cd3c352184ebffc1dfc1a8ec", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:44.884093", + "end_time_simulation": 9.979999542236328 + }, + "44c3b59edfcc41f494719efcfe482528": { + "parameters_hash": "24779a74058f32bda3d4313f30bbb6bc", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:36:25.021661", + "end_time_simulation": 9.979999542236328 + }, + "c7ee1c92985656d99843afbc206014c4": { + "parameters_hash": "9a31fe741cbb4d672812229e8c540327", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:36:28.192618", + "end_time_simulation": 9.979999542236328 + }, + "337efe3f652b4c12b24b70a63f25adff": { + "parameters_hash": "1f5aac387ceaacad16f1ba18eada37a4", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:35:47.782554", + "end_time_simulation": 9.979999542236328 + }, + "280e9d977d83502387ef8e05031424b9": { + "parameters_hash": "2c76c6467bac44ee9b5d4ee65d8773a9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:51.000761", + "end_time_simulation": 9.979999542236328 + }, + "322aa53bc0ae43c08d03196c3fa01dcf": { + "parameters_hash": "c2e3dade987a5c80aab3c25a1f5f0b21", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:35:53.362710", + "end_time_simulation": 9.979999542236328 + }, + "0888f32f92204a17a1936fef4dec8cee": { + "parameters_hash": "1c8ad6eddfabaa4b558a39f757df2e04", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:35:57.052178", + "end_time_simulation": 9.979999542236328 + }, + "9ab8a041936c48b1a3c9d3660f774e17": { + "parameters_hash": "9aa1221dc6ebb91cfdcf095f0a9c35a9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:35:59.836778", + "end_time_simulation": 9.979999542236328 + }, + "1c287818f1fb4ae9993f2132d42c76e4": { + "parameters_hash": "3c42b497e274f5e99a6995a70b07e7c1", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:36:02.347984", + "end_time_simulation": 9.979999542236328 + }, + "34d163353a5458bdb11ab2c068070160": { + "parameters_hash": "0155e7b31f658eeb959472cd3800ec06", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:04.748367", + "end_time_simulation": 9.979999542236328 + }, + "5ff590ebbeda4c21a919f7a6f15ecb84": { + "parameters_hash": "e0ecf8d4e72fc542c5ed84c70cce49b0", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:36:08.158083", + "end_time_simulation": 9.979999542236328 + }, + "e544644581fc43d2898f5895dfc1f942": { + "parameters_hash": "30d3ecdc996451c3dc35161baa9b3fc1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:10.751416", + "end_time_simulation": 9.979999542236328 + }, + "b880dea01a7a404fa6882f936385971e": { + "parameters_hash": "e572a4ca47335d3fe30f6a9136577fcc", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:36:13.648839", + "end_time_simulation": 9.979999542236328 + }, + "7571b485080d41bda21a8aea9b5a9288": { + "parameters_hash": "bb9bae62cca4851e6e9b07974d96c134", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:16.175819", + "end_time_simulation": 9.979999542236328 + }, + "2eaaaa5d6f8e4c1b8447b4389e52853e": { + "parameters_hash": "8eeca15cda7e4f0e8652c3ba0d704f71", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:40:01.402354", + "end_time_simulation": 9.979999542236328 + }, + "ad9d91ad86ed5581904a3c1cb882cc7b": { + "parameters_hash": "e1914754166c48789967bb6ef0220974", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:40:05.037024", + "end_time_simulation": 9.979999542236328 + }, + "9f631443f00d405b9541b1a883b22f08": { + "parameters_hash": "b4be2750bb1597165f81969f3c9460f6", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:36:19.055932", + "end_time_simulation": 9.979999542236328 + }, + "559a252c25414d139cb66654bb78fd25": { + "parameters_hash": "3b2fd0c793ed28af36866c114dcfd21a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:21.640992", + "end_time_simulation": 9.979999542236328 + }, + "825bb775cf154fdb983653ce20d329dc": { + "parameters_hash": "8034c71adb97fc649df2021cca124f95", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:36:31.838641", + "end_time_simulation": 9.979999542236328 + }, + "0d0f208513fb5a3b8664243d82c65687": { + "parameters_hash": "2f3e61cc19be1ffc6a90440d8ac01eaf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:36:34.204171", + "end_time_simulation": 9.979999542236328 + }, + "6557573a15fb4a4da5dc3929b4bf1eb6": { + "parameters_hash": "8753de8711c5081a1d1a442cb072f82b", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:36:24.429598", + "end_time_simulation": 9.979999542236328 + }, + "9918ef62c8bc5d4eb5a4410830fe71fe": { + "parameters_hash": "d6cc3db96722d1a4d3007234e67a582b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:27.362220", + "end_time_simulation": 9.979999542236328 + }, + "dd028d3ab7fb4aae84694d21ecf4c201": { + "parameters_hash": "70883e53be6c4efafeaa5db3677f2f9c", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:36:30.633963", + "end_time_simulation": 9.979999542236328 + }, + "900f3f2302884f96877e7c32afe81a21": { + "parameters_hash": "780a6a6b240bf4dc8ef5c1d9fd472a1f", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:36:33.579281", + "end_time_simulation": 9.979999542236328 + }, + "55bb234abc6f4c10b99d4a8aa8b6f081": { + "parameters_hash": "90245926bc2b50ae5a726026df3aff3e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:36.236348", + "end_time_simulation": 9.979999542236328 + }, + "b006ba4bcca74f5e991d209e8dc3b513": { + "parameters_hash": "78ecdae1cee8574b44a3299f53c014e0", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:36:39.637144", + "end_time_simulation": 9.979999542236328 + }, + "ed5f629d274f5ed2886bdc9bd4b27cc5": { + "parameters_hash": "062ea2be65e3f65e7db941e21be7df31", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:42.619191", + "end_time_simulation": 9.979999542236328 + }, + "526e572a5f59410384e5ea0861df7111": { + "parameters_hash": "253f024a5499c30ec74d14ccb0abf8a7", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:36:45.607714", + "end_time_simulation": 9.979999542236328 + }, + "0169e164efb5455e8d806b7011bed036": { + "parameters_hash": "8f5715315af385ca8dbbf39c9cd6a02e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:48.292888", + "end_time_simulation": 9.979999542236328 + }, + "d11428ddb1ee4631b1ae3661e2e4586b": { + "parameters_hash": "4689f5f30d8c8e6c999f7fac0ef82360", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:36:51.104021", + "end_time_simulation": 9.979999542236328 + }, + "6de4f83298464f27b1dd666491b1cee2": { + "parameters_hash": "d982506d22efcbfe0ec0037c4580f452", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:53.573769", + "end_time_simulation": 9.979999542236328 + }, + "88403ff5d33f40738c667d29ce40dfdf": { + "parameters_hash": "41a2cfcd4ed6407dd83e1642413fbccb", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:36:56.828596", + "end_time_simulation": 9.979999542236328 + }, + "5ffb39cbd48c4d39ad3bb152efe8f8ca": { + "parameters_hash": "4278c9760eac10be3d19eeffe028a11d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:36:59.433999", + "end_time_simulation": 9.979999542236328 + }, + "487567ed859142819ac3b3be79f2b15c": { + "parameters_hash": "ecc00ac760f225af80b7793c13d997c7", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:37:02.133168", + "end_time_simulation": 9.979999542236328 + }, + "9175d1d8f29248a0905d67b8348ea5d5": { + "parameters_hash": "dd835daaab93df64e59ae7b32f16e0d2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:37:04.769145", + "end_time_simulation": 9.979999542236328 + }, + "a4b8dcf0589e41e099065091244f12bd": { + "parameters_hash": "7bb1fb1470cee429b3d516b69b8f9f34", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:20:23.292902", + "end_time_simulation": 9.979999542236328 + }, + "a8db19a319ab5f9ea2affcc0b937cfc1": { + "parameters_hash": "ede44e25389ba54ca7dbf0b60f8950fb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:20:25.659652", + "end_time_simulation": 9.979999542236328 + }, + "eb4b41ba79394aa3b8057efd1fc99ac8": { + "parameters_hash": "bb210255031637e5dc1ea372db680a4d", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:36:37.852471", + "end_time_simulation": 9.979999542236328 + }, + "8c29819615f85e0e8f771708c9948239": { + "parameters_hash": "d79ffc7c1bd249f142ba4d7cb092ff59", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:36:40.600337", + "end_time_simulation": 9.979999542236328 + }, + "65255674cfb2409088660c4b9fac8bbe": { + "parameters_hash": "74b94948355f036907f4fdd88fcfb40d", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:37:07.780236", + "end_time_simulation": 9.979999542236328 + }, + "0be15e19e4cb46588d9dbcb22c154783": { + "parameters_hash": "e5fab8a86c1c1f561da66ea85d612a08", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:37:10.307295", + "end_time_simulation": 9.979999542236328 + }, + "bdf601681b8541659ad53f2b52d11526": { + "parameters_hash": "02fccb048ef7fb32d53361dc56ac8748", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T16:01:23.787430", + "end_time_simulation": 9.979999542236328 + }, + "ab1dbdc0c2344be7a4b105dd445c3e62": { + "parameters_hash": "6e06d292fc5f0e83e260d6805c3a3667", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T16:01:29.218019", + "end_time_simulation": 9.979999542236328 + }, + "e422c448fbee4a4e938550f40798e7cd": { + "parameters_hash": "221c48696bbc9a9e48a9f637638f0fc8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:01:32.332437", + "end_time_simulation": 9.979999542236328 + }, + "45b1dbc0577341cba4331b245b85e0d3": { + "parameters_hash": "525c6bc12223b8843d7d08846be04966", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T16:01:36.383927", + "end_time_simulation": 9.979999542236328 + }, + "dbc8ca063f5d511fae9210590e32304c": { + "parameters_hash": "4c3e9007dc736a5cd73f79787e09543d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:01:39.647035", + "end_time_simulation": 9.979999542236328 + }, + "97f285a3809940f68129ed3a9d1d9b26": { + "parameters_hash": "a30b5b00b5368b79cb13dbf2eae47b07", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T16:01:42.930689", + "end_time_simulation": 9.979999542236328 + }, + "f3086760d09042978e50763737d3d391": { + "parameters_hash": "b488cd9aeb6058b62abd13b1136fa4d0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:01:45.475014", + "end_time_simulation": 9.979999542236328 + }, + "b0b90fd4c4ff498b8adad05e6785c8ab": { + "parameters_hash": "c034cd7431168f56dd255404754b0f9d", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T16:01:48.983743", + "end_time_simulation": 9.979999542236328 + }, + "5ce7e26252f1419aa2e87a116335aa6b": { + "parameters_hash": "2a8c846bd43ce949a8e0e500d60674e1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:01:52.700295", + "end_time_simulation": 9.979999542236328 + }, + "1c5a4040d24e465689c3a47619c58be3": { + "parameters_hash": "d9bfa998cbc07ad9f9d7879f8142dd46", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T16:01:56.317191", + "end_time_simulation": 9.979999542236328 + }, + "678c6f56fa5a49aaade6aab2fbf27611": { + "parameters_hash": "0b05628db679cf5af9f307813fc8cdeb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:02:00.567326", + "end_time_simulation": 9.979999542236328 + }, + "c0bd708751dc4c758cc183aa2f80b147": { + "parameters_hash": "3c9a3b0384d2169430e4fab9a2cf2de7", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:02:05.047819", + "end_time_simulation": 9.979999542236328 + }, + "bf4168809ac258dd94001add2f7b43ca": { + "parameters_hash": "0fd7b52dc1f3f2fe607b27720c8b7cfa", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:02:08.155396", + "end_time_simulation": 9.979999542236328 + }, + "97c922e470a6420aaa0db38a7b47ab14": { + "parameters_hash": "cde69e48968965e5ffac6a34a8ae3532", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:02:10.997669", + "end_time_simulation": 9.979999542236328 + }, + "59ef06a1e6cb423eb2a43621204c161a": { + "parameters_hash": "973333a20c3e280ce7a3cd0a04a574d4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:02:13.740314", + "end_time_simulation": 9.979999542236328 + }, + "9c8def26273548e58053ebb6405233bf": { + "parameters_hash": "b1457fa8aa41f7635a6e0d528000a699", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:02:16.393343", + "end_time_simulation": 9.979999542236328 + }, + "ccaf317448074337942c21730f3cb2f1": { + "parameters_hash": "c33eead44b156c2972c198464ab1af4a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:02:19.488998", + "end_time_simulation": 9.979999542236328 + }, + "43a6a8ec17ae49b68a9ecd398c67f6a9": { + "parameters_hash": "ae97568431bf4507ca2936121c132d92", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:02:22.378270", + "end_time_simulation": 9.979999542236328 + }, + "43a4c50748a657e88e7ccbf3879c5a6f": { + "parameters_hash": "68969423cce81bb10288fa3e03b26a71", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:02:25.409616", + "end_time_simulation": 9.979999542236328 + }, + "852557979fcf490298f6359bfb41fbd3": { + "parameters_hash": "a76f9b97e467c76ae196d30f2e6c19aa", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:38:09.110114", + "end_time_simulation": 9.979999542236328 + }, + "93d61aa6d25243a6aaa6b4b8af95e653": { + "parameters_hash": "0b41bc8e64707a680c5915bdd1664a75", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:38:11.825262", + "end_time_simulation": 9.979999542236328 + }, + "b89de33f099843eba0876c611b07ec8d": { + "parameters_hash": "d217c770c873f6fccf72582f2ec06ea5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:38:14.590251", + "end_time_simulation": 9.979999542236328 + }, + "2f00e9a90e3c4484aa48107c021745aa": { + "parameters_hash": "f6a0a75c86fcd1fe59005083af35e754", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:38:17.597374", + "end_time_simulation": 9.979999542236328 + }, + "3d7f9936c68a5f7e8f8e3977a542cb3b": { + "parameters_hash": "1b137a502e67c62551457f197a8ff3db", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:38:20.286456", + "end_time_simulation": 9.979999542236328 + }, + "4704a3434062450f9cefc1dccc09379b": { + "parameters_hash": "a5677a2108e248978affa7e6321ec105", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:38:23.319954", + "end_time_simulation": 9.979999542236328 + }, + "04f1a97187ff471eacca2bbfb98fa36f": { + "parameters_hash": "190a67c3b955bf17c90e630636e20d61", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:38:26.101295", + "end_time_simulation": 9.979999542236328 + }, + "2ec3063d694146ac8feab2743f2d3c31": { + "parameters_hash": "24de26dda001eb3d1cb68a6bfea8edab", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:38:29.095494", + "end_time_simulation": 9.979999542236328 + }, + "b2fec1b4e60940148cd4098973907be3": { + "parameters_hash": "7ec2dd20280f41bf49750145e96fd102", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:38:32.070419", + "end_time_simulation": 9.979999542236328 + }, + "dc1d8c8ee6bf40e4b954f013f21f83f9": { + "parameters_hash": "d0a82db8fb33a84874e863aa77257c4f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:38:35.051435", + "end_time_simulation": 9.979999542236328 + }, + "d0bbe86dc6064e3993de291633b0bb5d": { + "parameters_hash": "4b7be37ac7a8c2cd93e950956a376f7a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:38:37.648496", + "end_time_simulation": 9.979999542236328 + }, + "c01d63d396c44b4eb0e0b49cf8b3b79a": { + "parameters_hash": "435ee7d8d5b7243cb1a611d27b67b9d4", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:40:09.330592", + "end_time_simulation": 9.979999542236328 + }, + "ced48cf503995ce3a0d097ed644d440c": { + "parameters_hash": "8415373f15bdc253fac50888be507c42", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:40:12.160984", + "end_time_simulation": 9.979999542236328 + }, + "07d05996d42242fcad5bc061474b54aa": { + "parameters_hash": "b0272a71a9f6f0627e5ad43ed447ca58", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:38:40.369943", + "end_time_simulation": 9.979999542236328 + }, + "22fcd6e8aee74bebb8109a9ede80e4d9": { + "parameters_hash": "33016549cd2ddfbe23be8bfcaeaf6601", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:38:42.538679", + "end_time_simulation": 9.979999542236328 + }, + "1937da4e366540ba887bff1fbf433de6": { + "parameters_hash": "4bf9675e996410a1fec090bcfc2e4c04", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:36:43.200694", + "end_time_simulation": 9.979999542236328 + }, + "7314650ae1d95e878464db310004d162": { + "parameters_hash": "90aa7c8c237dddb47e675b32e558de40", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:36:46.067494", + "end_time_simulation": 9.979999542236328 + }, + "03cf9a5fd2af43aa8b2413596e24ffa4": { + "parameters_hash": "911393fb9b44bf1e0ae603e8277066d8", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:38:45.332081", + "end_time_simulation": 9.979999542236328 + }, + "f9fa1adb91865ba4aa8c82092a56b4dd": { + "parameters_hash": "29658f3f6b7e06e03fe15e121eeec4fc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:38:48.007962", + "end_time_simulation": 9.979999542236328 + }, + "924138e0d0f9479f8dccc89cc12e589c": { + "parameters_hash": "fbdb5171316622c1063d9fba7c5b081c", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T15:56:02.622420", + "end_time_simulation": 9.979999542236328 + }, + "297d2782076c46f0ab96b3197aae195a": { + "parameters_hash": "d58fa5ffdc5a9a4863a9e8850e594f97", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T15:56:08.520990", + "end_time_simulation": 9.979999542236328 + }, + "17ed1d6bb5a94f958d6593efd77b8712": { + "parameters_hash": "31f8dc52f0042ff070f238b642fb6baa", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:12.348965", + "end_time_simulation": 9.979999542236328 + }, + "3c9308c66d644b07be5b2c3ee6ffa4e6": { + "parameters_hash": "6e8d970bba6c0c8952eda0ab9cd8cd72", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T15:56:16.105288", + "end_time_simulation": 9.979999542236328 + }, + "64cadb25079a5209aa3b1a66b5141546": { + "parameters_hash": "3f49ec046465aeb3af321c7898e1b387", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:20.505347", + "end_time_simulation": 9.979999542236328 + }, + "41f6d71366a248b18e4f1bd2e9e837fc": { + "parameters_hash": "f06c1ab7ad3ee1aa9ccad7c642891eb9", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T15:56:24.178725", + "end_time_simulation": 9.979999542236328 + }, + "0882b2cb7fb1591ba9ed15d317a6cb17": { + "parameters_hash": "8dd8b4ae75ee7fbc19115db772a08f1c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:27.626599", + "end_time_simulation": 9.979999542236328 + }, + "365a77e2d8044c9cab8a8d5d9fb7c3be": { + "parameters_hash": "fba3527c95f6fd8fa3ce042ea47cc1b8", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T15:56:31.614370", + "end_time_simulation": 9.979999542236328 + }, + "9f783f69955e476b97c8c3f8ad33b02f": { + "parameters_hash": "ed8cb960bff69b3bc80bf6479b8b5d03", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:34.674188", + "end_time_simulation": 9.979999542236328 + }, + "b0157d3a8b564b1fb882e4b145411c7a": { + "parameters_hash": "21689c7ad77d0154f54df445be804f68", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:56:37.380868", + "end_time_simulation": 9.979999542236328 + }, + "bd6e1dd535ff5368aef8961b1d508a3a": { + "parameters_hash": "2f5173de0de6b6ad22b8c9bbd072d8e6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:40.017928", + "end_time_simulation": 9.979999542236328 + }, + "5e751fe95803410a9e030d94147883be": { + "parameters_hash": "a4bd0cfe8393900f4050732f495f67ef", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:56:42.436671", + "end_time_simulation": 9.979999542236328 + }, + "f8fb591a28f3534ab0a4039429f586b1": { + "parameters_hash": "6ea3fba132e405ca664cd987df42f777", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:45.301711", + "end_time_simulation": 9.979999542236328 + }, + "92899d2064ba4768ad20c3c4e2e26102": { + "parameters_hash": "8cd01f99fefd9bcd3fadad8f8ddb5275", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:56:48.176388", + "end_time_simulation": 9.979999542236328 + }, + "94516dccf4fd4871b891831932e94982": { + "parameters_hash": "1393c474c96ed92e01ce061061dac0a3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:50.348050", + "end_time_simulation": 9.979999542236328 + }, + "9e10b79464c24f36a53f252be6541d5f": { + "parameters_hash": "3a6ea1d1c0a69068dcc5c61e6f645066", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:56:52.729973", + "end_time_simulation": 9.979999542236328 + }, + "ba55963ceb564eeea8921c6cd5621e31": { + "parameters_hash": "bcb90856191c57371092f10d29566e66", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:56:54.776486", + "end_time_simulation": 9.979999542236328 + }, + "afec7fdef34d4256a0060406c9b1cd7c": { + "parameters_hash": "99654a2b38c486d209c90d8a42d4e431", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:39:36.992961", + "end_time_simulation": 9.979999542236328 + }, + "9b070abe58ea4728b174e46b35a38eff": { + "parameters_hash": "57d126bdcd5bc2a40509990a9a3d8e7b", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:39:40.566553", + "end_time_simulation": 9.979999542236328 + }, + "353418978d9640ee904556c7b46dfc6e": { + "parameters_hash": "0aad1dba30a66842e5739a6836c6dfb0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:39:43.600528", + "end_time_simulation": 9.979999542236328 + }, + "b905c7fef87e4494a9e3ef708768bda9": { + "parameters_hash": "bc34f59ec8ccc8ca6a7aa243ac86567a", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:39:46.997604", + "end_time_simulation": 9.979999542236328 + }, + "dc0673001a4b5fb98ca0c4ca0e0660dc": { + "parameters_hash": "49910eecb5d9dedc5ac8bddbd620ead1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:39:49.256678", + "end_time_simulation": 9.979999542236328 + }, + "260750f9d27a4578ba899fcfb02ac653": { + "parameters_hash": "47f716a3f81a37b815ac8edd619a0ceb", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:39:52.649211", + "end_time_simulation": 9.979999542236328 + }, + "59e7b0e654f5403195a7b86bb2aef5c3": { + "parameters_hash": "b5456ba225285650760d51093ea53f45", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:39:55.151806", + "end_time_simulation": 9.979999542236328 + }, + "b9f30d1a0bc8496f9c81a385b4fe34f6": { + "parameters_hash": "221621c9ab579fd6618ffb4b84a55cde", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:39:58.062769", + "end_time_simulation": 9.979999542236328 + }, + "3b7073f80f8e473cb14a1a75d35d653c": { + "parameters_hash": "ac7b6ac6abc67732a21c21600a63f838", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:40:01.705572", + "end_time_simulation": 9.979999542236328 + }, + "d1713cff0a254d2e9079c72c7cb32e46": { + "parameters_hash": "86033e539932f21962d0c17ad16e90f8", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:40:05.386325", + "end_time_simulation": 9.979999542236328 + }, + "9544f57b9ef0545bb441ade3a1e51f48": { + "parameters_hash": "984d3022719974ea77026e4db2cac2e9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:40:08.597803", + "end_time_simulation": 9.979999542236328 + }, + "3a518eedec2f4abba80d37e07b821eb9": { + "parameters_hash": "735da05a3f3739e0f15dc1ec27e1014e", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:40:12.268291", + "end_time_simulation": 9.979999542236328 + }, + "2c2335cd32af446aa2821844ef71349f": { + "parameters_hash": "91b428fdfe88235915ab7e987f89bb7f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:40:15.230983", + "end_time_simulation": 9.979999542236328 + }, + "491d2f58388f4951952ab7972d72ad3e": { + "parameters_hash": "0ffd6189e422dbc70a51c0e507640552", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:40:06.991111", + "end_time_simulation": 9.979999542236328 + }, + "80ec333f5e434ec89f1e27e27caa868a": { + "parameters_hash": "0219872fd4f01af0f5308c19aa521b2c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:40:09.922634", + "end_time_simulation": 9.979999542236328 + }, + "ffd5c2759ad34607be8666b859ae6bc4": { + "parameters_hash": "1d91a588d026605315175ed638fcd683", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:40:12.651825", + "end_time_simulation": 9.979999542236328 + }, + "545efe321e5e549e944307e764c9b00c": { + "parameters_hash": "b94929aaa9be3e4143c9c8b74c94da94", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:40:15.367795", + "end_time_simulation": 9.979999542236328 + }, + "cac45b19acaf43b190ce566ea5b23d9a": { + "parameters_hash": "0d32a83a315a5459316036208fc542a0", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T19:31:23.604479", + "end_time_simulation": 9.979999542236328 + }, + "f2f945d1aa78445c80b540f6b650a6d5": { + "parameters_hash": "01dbb874032e523972a05a1dfb4789d0", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T19:31:29.645631", + "end_time_simulation": 9.979999542236328 + }, + "81b2921074c64cc790c424376b78045a": { + "parameters_hash": "a911b8d051f41ce7d4970017e9ff23e3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:31:33.265792", + "end_time_simulation": 9.979999542236328 + }, + "776a7eb5016b4ab482db69657bac228c": { + "parameters_hash": "1d1d3e0a62a4564bcaef62678c556cd6", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T19:31:36.121367", + "end_time_simulation": 9.979999542236328 + }, + "8f7483093e245c7d9de0bc61fbf7c7d7": { + "parameters_hash": "4f3e359bd81b53db58f6d57d10bdc42d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:31:39.036433", + "end_time_simulation": 9.979999542236328 + }, + "fb19e96e47e94589b918ed609aee4062": { + "parameters_hash": "9e12e4d6a506d76002bececbd778af1e", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T19:31:42.566884", + "end_time_simulation": 9.979999542236328 + }, + "feba64e8585c417db3e0710b9f70b76f": { + "parameters_hash": "64258a332e1989a1c0730ba3e5125c0a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:31:44.863479", + "end_time_simulation": 9.979999542236328 + }, + "5c6ece223a1a46f6a1987449154ff138": { + "parameters_hash": "72ae9881ceb2f8e55efa36fe0d6cb64c", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T19:31:47.186050", + "end_time_simulation": 9.979999542236328 + }, + "8a698338f227525da2470df73f537581": { + "parameters_hash": "21c830deba2cbbcf722d2a4def135584", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:31:49.469078", + "end_time_simulation": 9.979999542236328 + }, + "35da4fcd58cc487a8e0125c2372e5e95": { + "parameters_hash": "cf41a6ddd720c84b6b7ca493e2d99dfb", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T19:31:52.185831", + "end_time_simulation": 9.979999542236328 + }, + "d10362314f4443adb722048f9dc07a4f": { + "parameters_hash": "b6b66d4f686c27edb30950b842b75662", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:31:54.472505", + "end_time_simulation": 9.979999542236328 + }, + "c405fc845cb24b21a0541d929922f563": { + "parameters_hash": "d115e8a30335c1d7ca84713b7f4c3769", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "failed", + "timestamp": "2026-05-05T19:21:42.547033", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "fbb37dd51339582da246bc6607801e3e": { + "parameters_hash": "1db2d4a32cb0d58b889c8f90b6a268e0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:21:44.265989", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n" + }, + "38a41ef6d3a740d78b643308404c3754": { + "parameters_hash": "9e316317b7f26ac0b59e6b6a9b970071", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:31:57.206990", + "end_time_simulation": 9.979999542236328 + }, + "eb4833ebdde748ab9ced3f3f3f0e6b36": { + "parameters_hash": "c0f72d9d961622279dbe9b8ed76e9975", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:31:59.226276", + "end_time_simulation": 9.979999542236328 + }, + "f3c513de8c2d4271bd85479cd73fee3c": { + "parameters_hash": "0c27fc02499410a4c97b08aa0041fa68", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:36:48.919499", + "end_time_simulation": 9.979999542236328 + }, + "0f36071da0535705a77da10d686075b0": { + "parameters_hash": "864900db6637fa8fed763461b4dd3438", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:36:52.462209", + "end_time_simulation": 9.979999542236328 + }, + "5756fac85c144250a3f4b38c30340577": { + "parameters_hash": "89b76034e4117c3757666bb89488a294", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:49:02.779468", + "end_time_simulation": 9.979999542236328 + }, + "60c933a58d2659ae8f25a6fc099a9e65": { + "parameters_hash": "65518fd552bc62f53cc9d51eb218a420", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:49:06.266793", + "end_time_simulation": 9.979999542236328 + }, + "4f105c22a575422faa161016192329d2": { + "parameters_hash": "7e985b840cd364eb6263ca8a1d070f39", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:41:15.662830", + "end_time_simulation": 9.979999542236328 + }, + "191103819e8c4df6be9deb32e72b4d50": { + "parameters_hash": "f5fee7f2ff037e7b0abf6ea37d2939d8", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:41:18.571861", + "end_time_simulation": 9.979999542236328 + }, + "224290ce8932493c8fd221fa6fac5623": { + "parameters_hash": "fe545f6d7f0568bbb9b20876055a9b89", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:20.913608", + "end_time_simulation": 9.979999542236328 + }, + "3d97964fb7c24cc291fbc25dea19c9d4": { + "parameters_hash": "be56fc0b8e91df425c208d2fc13d7349", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:41:23.386084", + "end_time_simulation": 9.979999542236328 + }, + "bc1636fd7e795e3ba88f81628071320a": { + "parameters_hash": "3da7e3a090630fce59238c00f393191b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:25.934976", + "end_time_simulation": 9.979999542236328 + }, + "58836b2e1ac24cddb0349f7a5ef30cdd": { + "parameters_hash": "f3cc3bebecda33305824fa0c9f1ab856", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:41:28.949340", + "end_time_simulation": 9.979999542236328 + }, + "5b274ae8d9cc434e921c5f669c8cc57d": { + "parameters_hash": "93d2d5ee37c331fd886fa9b26e40a773", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:31.221523", + "end_time_simulation": 9.979999542236328 + }, + "4f9403152d1f44e798264781c4aa5389": { + "parameters_hash": "06a13c2b3a21f79cd50df99be3e4d40b", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:41:33.454861", + "end_time_simulation": 9.979999542236328 + }, + "def6065a67e4473f9df3fba117125390": { + "parameters_hash": "6f83043c26c479cc92c337431a6f45dc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:36.285635", + "end_time_simulation": 9.979999542236328 + }, + "477f6e75c9c24d15b8bde23696df914c": { + "parameters_hash": "cf880be986e2f71442ae39cb7f2633fc", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:41:38.851118", + "end_time_simulation": 9.979999542236328 + }, + "09d1a966e05445588b76810e4894a918": { + "parameters_hash": "36599ed5bfaf851abbafdf9fb9b3984c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:40.823410", + "end_time_simulation": 9.979999542236328 + }, + "cd5b8bdef4d44b7d8f6682e3cde5fcf2": { + "parameters_hash": "3087b5c9008db5fa7b442d4d329c21dc", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:41:43.701562", + "end_time_simulation": 9.979999542236328 + }, + "e6851d0304744cd5a5946e56b680d18b": { + "parameters_hash": "17b8538c13542707e6343fe1dc7748a8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:46.100440", + "end_time_simulation": 9.979999542236328 + }, + "1d716244585943cc9e58b841706c0a36": { + "parameters_hash": "116e6256a44c78a215649d8424c764ba", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:20:56.442450", + "end_time_simulation": 9.979999542236328 + }, + "e9505f333e335987b48984750bcccd89": { + "parameters_hash": "0e79d2b1b78a3076610ae2cfcee75b6c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:20:58.541880", + "end_time_simulation": 9.979999542236328 + }, + "153a26e412554d13b5edd43184000e53": { + "parameters_hash": "bf3058e01de7486e518037f90200beb7", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:36:55.478726", + "end_time_simulation": 9.979999542236328 + }, + "d122354944d05990b37407f237d5c54d": { + "parameters_hash": "8b6ffcb6446a164a23a45db3acc9e028", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:36:58.060590", + "end_time_simulation": 9.979999542236328 + }, + "6b927147cb424095a7f44176813983bb": { + "parameters_hash": "5b65e2d217c64341ceee3f40d1f0dbef", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:41:48.783823", + "end_time_simulation": 9.979999542236328 + }, + "ccfbf91e4a254a18900d7a15d59af024": { + "parameters_hash": "6240bad4541204f5375497ad6c5752f6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:50.933807", + "end_time_simulation": 9.979999542236328 + }, + "96814fa562ac4bac8cdd1ace014be742": { + "parameters_hash": "093451c4ffe836f6991a0837aac5b941", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:41:53.348367", + "end_time_simulation": 9.979999542236328 + }, + "e430606893ee431087b189d71cd29ac5": { + "parameters_hash": "407143b8a9c3bffbb0212727c69f68e0", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:41:56.137911", + "end_time_simulation": 9.979999542236328 + }, + "13304d23a4aa4ac5b2304a58f9d8207d": { + "parameters_hash": "fc0017f1db7e2bb830a4216fab3f65c5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:41:58.391410", + "end_time_simulation": 9.979999542236328 + }, + "bef6610a0f3e46ecbdbbdeae71e18a67": { + "parameters_hash": "afbc65deddef1ccbd516b4f50e6a936d", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:42:00.647175", + "end_time_simulation": 9.979999542236328 + }, + "c4a241c2ee795aaa8380227dcd4eadde": { + "parameters_hash": "f48a2b5b73dfca595bb082cdfd4ead22", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:02.650506", + "end_time_simulation": 9.979999542236328 + }, + "9b572270b12942a390d67ff9595a2cbd": { + "parameters_hash": "5eec7d64ce95c19cce60da50fbac57b1", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:42:04.794649", + "end_time_simulation": 9.979999542236328 + }, + "d5de5d83796746968560827a30e73982": { + "parameters_hash": "29927bad62181490e2bc0680bdefbd23", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:07.219296", + "end_time_simulation": 9.979999542236328 + }, + "0429ebe0904a49e6a4cf24235e8f4358": { + "parameters_hash": "e8368fcde2ab323ea26ed37d62ddf7d4", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:42:09.768567", + "end_time_simulation": 9.979999542236328 + }, + "5f385b6676a648eda28391fbd2777e2a": { + "parameters_hash": "cd2391a0c08c4e7e78aeb99fa5ab2619", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:12.118689", + "end_time_simulation": 9.979999542236328 + }, + "7f628146eac84b60bfe2378bc6fed300": { + "parameters_hash": "46ad8852ec9325b28e51e7648866c067", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:42:14.899854", + "end_time_simulation": 9.979999542236328 + }, + "913e4d9972474404a1a5c7bf40c7db49": { + "parameters_hash": "685b9cc544fb9c3a3741faf332358343", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:17.238955", + "end_time_simulation": 9.979999542236328 + }, + "87f44ab26a9a410e9d59171eb1eb2a19": { + "parameters_hash": "b618521babfca5483c3709383d98cd95", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:42:19.738037", + "end_time_simulation": 9.979999542236328 + }, + "9323986239035b919328bb901225a3a3": { + "parameters_hash": "1761c53db89bdadfda57ed0224da2b35", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:21.907515", + "end_time_simulation": 9.979999542236328 + }, + "badd89b9580a4d319ec3f8a0bee814b1": { + "parameters_hash": "6f47662764e804f36d5be292e31099eb", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:42:24.380388", + "end_time_simulation": 9.979999542236328 + }, + "34230065d3b04cee9d391d9456fb8e2f": { + "parameters_hash": "54dc22247fdb030a1b187cb666ef1441", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:27.058614", + "end_time_simulation": 9.979999542236328 + }, + "99c66580a6ce47439d9640981d2c100d": { + "parameters_hash": "8e8f4ce50a96a1aa288fe8b37126dec0", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:42:29.867933", + "end_time_simulation": 9.979999542236328 + }, + "5e88c9043d6e40ab8fae031a0575c255": { + "parameters_hash": "430c62c496d5f0f21ed3d8f13d493918", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:32.862240", + "end_time_simulation": 9.979999542236328 + }, + "a3bc5bf3fff0436593782dba6e75baa3": { + "parameters_hash": "05b8a9d5db0ecb098a8a606dea23563f", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:42:35.346327", + "end_time_simulation": 9.979999542236328 + }, + "3fd1479581a258d59e445c04448cb612": { + "parameters_hash": "74dc4640bee129bfdd584801d28877d5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:38.427499", + "end_time_simulation": 9.979999542236328 + }, + "b6e5860643684682b60e6c146121b2be": { + "parameters_hash": "2c5d53ece8f1fa28d30ef0ce13024768", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:42:41.077114", + "end_time_simulation": 9.979999542236328 + }, + "833168b23c474c3f95c1ba4d9fff6176": { + "parameters_hash": "425f8e0929b7fdddec4c7aa6982338db", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:42:44.008289", + "end_time_simulation": 9.979999542236328 + }, + "33fbe4ad34534e04832396da0257093b": { + "parameters_hash": "0d5b4f54f8de23c3ae5a4209912fbf9b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:46.920168", + "end_time_simulation": 9.979999542236328 + }, + "b5518f5e667f490594eddd0536856609": { + "parameters_hash": "8751fa41ab06534eec2db1083fdac856", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:42:49.846025", + "end_time_simulation": 9.979999542236328 + }, + "bbe7fb968b115cb39ea2b4ec9099ff40": { + "parameters_hash": "ff36f2d8422e4835341f7538f50febd0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:52.722357", + "end_time_simulation": 9.979999542236328 + }, + "78b58040b64b4752a8c3bfe5b599fbba": { + "parameters_hash": "e53769d90b847db19fe7cff4eacfd4fc", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:42:56.565873", + "end_time_simulation": 9.979999542236328 + }, + "3aa854ab08d945328f5be100e17303b6": { + "parameters_hash": "1bde50fea9f26e19467b16e67e1a21ff", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:42:59.483275", + "end_time_simulation": 9.979999542236328 + }, + "46a0e4d74477440fbe3a46eecede6449": { + "parameters_hash": "589768d2c95d95ae9b204a79fe46f6f4", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:43:02.506923", + "end_time_simulation": 9.979999542236328 + }, + "f5e45b86eee640b79ebc25bff575cc68": { + "parameters_hash": "f4967a72cd1bfcfd4c32f99e6f8771b8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:05.287984", + "end_time_simulation": 9.979999542236328 + }, + "b2ab0497a7e541e69eae763994f29508": { + "parameters_hash": "1c514c3be720dbb85d6920f5accffebe", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:43:08.791043", + "end_time_simulation": 9.979999542236328 + }, + "7f33edac50a04f2d854b35cabeaa4b52": { + "parameters_hash": "d614a39789f1c3c02ddc407eef7c1e80", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:11.535334", + "end_time_simulation": 9.979999542236328 + }, + "bd1db72c3cdb481caa778ac3639cccb3": { + "parameters_hash": "2830941891adeffb526c72a63271f944", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:40:28.280021", + "end_time_simulation": 9.979999542236328 + }, + "0e49eba8274f5d02a8c8d842f048df34": { + "parameters_hash": "3b463b5a8fc03a6f751e42157c3aa118", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:40:30.415735", + "end_time_simulation": 9.979999542236328 + }, + "6c36644f43d64ade808ae9da3531c024": { + "parameters_hash": "2cdcc20af7e023e23fbb4bd7f76d7fd9", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:43:14.075079", + "end_time_simulation": 9.979999542236328 + }, + "b92113a134c243288b69b725ce4de7fa": { + "parameters_hash": "424bc27622a432ba4c5ea4282a7331a4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:17.172185", + "end_time_simulation": 9.979999542236328 + }, + "9829e3384fae4a77ae3bc21c6c8a6aea": { + "parameters_hash": "ea935ea072de6c0dd32b104da965a06b", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:37:01.077253", + "end_time_simulation": 9.979999542236328 + }, + "46c9bcce27e25c93bf330e0964fc2c70": { + "parameters_hash": "233d0d39510c07836177329402cc468b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:37:03.421850", + "end_time_simulation": 9.979999542236328 + }, + "1aa09eeb0ee541a0b2d9e4dd5e090b36": { + "parameters_hash": "10d041d89b60686e307f1b6459d19024", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:43:19.982765", + "end_time_simulation": 9.979999542236328 + }, + "e6d783268ec85bda8fa24a482cba0f3b": { + "parameters_hash": "fccac154541153c19d792c5991f78d38", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:22.103866", + "end_time_simulation": 9.979999542236328 + }, + "2713905d0dfa4d56aac01b96531ab3e7": { + "parameters_hash": "1c34914631acc2b73a0f8a8f473d4557", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-03T20:52:22.141306", + "end_time_simulation": 9.979999542236328 + }, + "62f46075a411443ca55bdec3e378fa61": { + "parameters_hash": "a0f12c28ac706b6559f7946ccbe69cea", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-03T20:53:35.426210", + "end_time_simulation": 9.979999542236328 + }, + "81c52b221d554d0292192fdd31f5d410": { + "parameters_hash": "5cb4c07ec913601dfd2e68dd2b4e03e8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-03T20:54:43.906347", + "end_time_simulation": 9.979999542236328 + }, + "c6fe6ed809c1448e8a94e437bdc2c44c": { + "parameters_hash": "b28fda80c45054bf31f66aa467efca62", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:43:34.677130", + "end_time_simulation": 9.979999542236328 + }, + "2b67ba0c1ea55eb3b4a5d9d5f0011b6f": { + "parameters_hash": "2aa5a9ec713f9e0633f8e1ce2fd2e3a1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:37.215021", + "end_time_simulation": 9.979999542236328 + }, + "223c934b74a24e5f9df19997035eef2f": { + "parameters_hash": "006d163d2ab583e31a904123fdd046d0", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:43:40.889959", + "end_time_simulation": 9.979999542236328 + }, + "f8c740dfaa764cce8ed490d6d4afa556": { + "parameters_hash": "d0220f330593058ab91c728838554c3a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:43.441658", + "end_time_simulation": 9.979999542236328 + }, + "72a49643adf94ae5b054b04699229c34": { + "parameters_hash": "7cb282a34d2ee1ca5a1db7705db6de86", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:43:46.242139", + "end_time_simulation": 9.979999542236328 + }, + "27ea42f0aa0745af94fb75825759cb15": { + "parameters_hash": "c1201def79d9f35cfc77c8ba26e40c21", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:48.892040", + "end_time_simulation": 9.979999542236328 + }, + "4f5bfa8cdd704608a47a35685f1e5295": { + "parameters_hash": "fed102187722b69d4faeebc015d71a8e", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:43:51.542165", + "end_time_simulation": 9.979999542236328 + }, + "f9c566a21b0643dbacadb6b69e6c120e": { + "parameters_hash": "ed1d685158bbfb7d4155f89d288ed1ac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:54.033434", + "end_time_simulation": 9.979999542236328 + }, + "836acbc3bc7c474398390e7ad0ff2d77": { + "parameters_hash": "eccc457f2902fb86d72015f9ddcfecd5", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:43:56.765380", + "end_time_simulation": 9.979999542236328 + }, + "1e3628363c3e456b84e9453e74c64143": { + "parameters_hash": "4a13aa205d50644022d9ae5443ff798f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:43:59.157110", + "end_time_simulation": 9.979999542236328 + }, + "a788bcbddb3e4143afc049ae6fd613ed": { + "parameters_hash": "a4cbaac4e7dd1e19fe2bb5c41a814c2c", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:44:02.301241", + "end_time_simulation": 9.979999542236328 + }, + "c3d3d75e9a1a57c1b2ad9c4c40fb58db": { + "parameters_hash": "fc22956daecca8311583717b36d7b0ea", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:05.833303", + "end_time_simulation": 9.979999542236328 + }, + "7b689cd67c2d4021930e364b1c2ed8ca": { + "parameters_hash": "aa8b6366a6896dd0bba26cb029df2e33", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:21:14.231397", + "end_time_simulation": 9.979999542236328 + }, + "638f59f63eed589a997a7c3a7af94479": { + "parameters_hash": "8a5effa184f49a328e44d796a4d6bdfb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:21:16.330536", + "end_time_simulation": 9.979999542236328 + }, + "656dfd2baea34c3182437fd93d50c277": { + "parameters_hash": "d8ca35d7be5c80e2dbc0b9c79660d7a1", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:44:09.037296", + "end_time_simulation": 9.979999542236328 + }, + "a1416e3200a6485294ab31cfbdcf30b2": { + "parameters_hash": "637baddcb2f4dad919f4604157a1c370", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:12.206949", + "end_time_simulation": 9.979999542236328 + }, + "d174df0577774f00a8ee5696da18039d": { + "parameters_hash": "f90ab1b3977f98a2e468e4bf7c94e7f2", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:44:14.606398", + "end_time_simulation": 9.979999542236328 + }, + "58837535fd5041a19b3fdd42bbea4d6b": { + "parameters_hash": "b2cf588653deb4317031ffee85c086a3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:17.303335", + "end_time_simulation": 9.979999542236328 + }, + "e4faa5a0e1ad4ca4b94bf2f18f4a093f": { + "parameters_hash": "385969c2db1796fe0788c245f81ea1ea", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:44:20.094844", + "end_time_simulation": 9.979999542236328 + }, + "4cca69d0d8004e33a522413ab396773b": { + "parameters_hash": "3839e3807aa0e8d340ed88d1dcb14e33", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:44:23.241556", + "end_time_simulation": 9.979999542236328 + }, + "8e9afe81a6d5457eaaf28f6891aa0fde": { + "parameters_hash": "15f39bfca8cc5a2436c1a44f66d1e696", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:26.135416", + "end_time_simulation": 9.979999542236328 + }, + "6fda857d20574e5aa143bffd1f17556b": { + "parameters_hash": "c9af15a8f80978b5a3841f836035fb44", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:44:29.869374", + "end_time_simulation": 9.979999542236328 + }, + "4f176e7e996951c8850959e74f7e9e1b": { + "parameters_hash": "164a80ae6349eeb0439224183f5b10a9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:32.659800", + "end_time_simulation": 9.979999542236328 + }, + "d1141599d62d4cd8a57708873e904583": { + "parameters_hash": "1f27aef26a69058a1f42a08e58f3dda6", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:44:36.029302", + "end_time_simulation": 9.979999542236328 + }, + "9d4e7802ebed4a5290036238f658beff": { + "parameters_hash": "1cd18151bba3a5be9fc63524678f7da6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:38.949442", + "end_time_simulation": 9.979999542236328 + }, + "b6a00423859b48e8b634aaf7a4b4208a": { + "parameters_hash": "0775335a3ecb8e62c3eb76f6101a1815", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:44:41.863619", + "end_time_simulation": 9.979999542236328 + }, + "aae3bea6ac124326b2331674e6c586b7": { + "parameters_hash": "a15c5bb050ac7aa5e8f38db72d394533", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:45.496496", + "end_time_simulation": 9.979999542236328 + }, + "7d482a4fa078411cb6b20df2b5319a7c": { + "parameters_hash": "4fb8b1daff6e441305308acfae170da9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:44:48.940263", + "end_time_simulation": 9.979999542236328 + }, + "2b02a99919584c0d927b2861586fda62": { + "parameters_hash": "bb0eb224a359fc9ec0b3c478007249d5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:51.575589", + "end_time_simulation": 9.979999542236328 + }, + "75878b62e856429ba0f400ca89099bc6": { + "parameters_hash": "c859c743a784c9abb15267c2e3398886", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:44:53.999487", + "end_time_simulation": 9.979999542236328 + }, + "7bac864a5196481bbe0255798182c2ab": { + "parameters_hash": "3a1790eb6408eba418e3de031f1f0038", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:44:57.540115", + "end_time_simulation": 9.979999542236328 + }, + "8b14f423b57a4377a418ca7fa2b5945e": { + "parameters_hash": "2271bf593731f84a10dd85b4cb811039", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:45:00.110406", + "end_time_simulation": 9.979999542236328 + }, + "9ef637c5959e5641b5d909b752c8947e": { + "parameters_hash": "135263dcdbfe29a8c6c70b4a2958b4d8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:02.567658", + "end_time_simulation": 9.979999542236328 + }, + "208ef509e8b04166ae5333a2877f1295": { + "parameters_hash": "a44a8f5df70186b459b4a2564f89e367", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:45:06.374628", + "end_time_simulation": 9.979999542236328 + }, + "0de2f0bb453249689ea1ac99a86bab8c": { + "parameters_hash": "0d54ecd4b100c5effbfe8c6037cd5bc6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:08.942112", + "end_time_simulation": 9.979999542236328 + }, + "12f391e2705a4e07965302c50b2efa0a": { + "parameters_hash": "9fcde96990b7af8322f6858001cab3a2", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:45:11.745331", + "end_time_simulation": 9.979999542236328 + }, + "39aeb78696124232bcec2e0fbd522a50": { + "parameters_hash": "dab494110b578ccb52eae7b432743e1a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:14.746078", + "end_time_simulation": 9.979999542236328 + }, + "b036100990f04a56b6e86b7ff9a5b948": { + "parameters_hash": "aaf5dbfe0ca18be57d0b82389b38dc06", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:45:17.988999", + "end_time_simulation": 9.979999542236328 + }, + "fb30c9848ce55947a903196f0d8d3aee": { + "parameters_hash": "d9dcbd3d34237c8e9e5b791655236d3e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:20.747301", + "end_time_simulation": 9.979999542236328 + }, + "914b438477dc4740af3fa7291646e9ce": { + "parameters_hash": "6716c712b28e55bbd1f3dacd1c1ae1c9", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:45:23.650862", + "end_time_simulation": 9.979999542236328 + }, + "450660a3c2d24f5bae3eeec80638199c": { + "parameters_hash": "0278fcdfaf128057578d409d99cc85de", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:45:26.691334", + "end_time_simulation": 9.979999542236328 + }, + "3906a5b3af3e43079b22aecd2305b089": { + "parameters_hash": "baff1a1cd1e9d9737d744628dce768c2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:28.841871", + "end_time_simulation": 9.979999542236328 + }, + "59b2dd9b949c42e7ab404612b867f051": { + "parameters_hash": "85e2764df08019ea6685e06982ba4b4f", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:45:31.887014", + "end_time_simulation": 9.979999542236328 + }, + "35e1bbf4dce6517da70442a7d9f6f982": { + "parameters_hash": "9d433d6527bfac496471f64be6221ce3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:35.203231", + "end_time_simulation": 9.979999542236328 + }, + "0a64f2cc7a1842988e0b2943c9c12fa8": { + "parameters_hash": "99a8643abc0ce5ab5546f34cdd71979a", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:45:37.845844", + "end_time_simulation": 9.979999542236328 + }, + "72fd75d284be423bbfaaca999b7368c4": { + "parameters_hash": "62210556effc6c3702309afbc0f37cad", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:40.017957", + "end_time_simulation": 9.979999542236328 + }, + "c79674d8f5aa438cba91cd6d92e57061": { + "parameters_hash": "960bb14c9e4f50e643e081d3324344cd", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:45:43.161582", + "end_time_simulation": 9.979999542236328 + }, + "1870663e97d84eb19ebd2a7c0f38b99e": { + "parameters_hash": "3e90798a64ec759e152945dc0a3d72e3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:45.763002", + "end_time_simulation": 9.979999542236328 + }, + "b8d4b4803cd242b19fd7a36747f378f4": { + "parameters_hash": "1ce07daecc35669e4e6e61f4b222dc33", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:45:48.691633", + "end_time_simulation": 9.979999542236328 + }, + "597e7f9a8ec449629e35e56771ef096d": { + "parameters_hash": "f8af32cc45c4985cf44c2a16e74262ae", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:51.196278", + "end_time_simulation": 9.979999542236328 + }, + "a97492441edf4e908ec9f55068fd1c4e": { + "parameters_hash": "cfb988cfc17c9b2f62db6434eaa24a7b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:45:54.262648", + "end_time_simulation": 9.979999542236328 + }, + "dbe8eb49d4b84eca81dc753a5acde3d2": { + "parameters_hash": "9a46a0a575654bc93e17d2459705b2b9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:45:56.687727", + "end_time_simulation": 9.979999542236328 + }, + "5ce98a8a2f3649c585b8afca8aa9bff8": { + "parameters_hash": "66bdd3fd9d444ea5a31e7f6143db7aa7", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "failed", + "timestamp": "2026-05-05T19:21:58.073932", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "4be3786039f65607908b54586a590598": { + "parameters_hash": "0f33e07783571010f778eb82a4c7bf88", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:21:59.389045", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n" + }, + "e3d9cf1c36474964bb478cb8b7f96015": { + "parameters_hash": "43f1eb8d32f2e29bca1cac6f47bd6d4b", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:45:59.899079", + "end_time_simulation": 9.979999542236328 + }, + "ceaedc9ca6834f7989eec19e1ccf2182": { + "parameters_hash": "c4f0bec1118074fabd2818a67af894e3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:02.356938", + "end_time_simulation": 9.979999542236328 + }, + "c228234aa2b443bbbbe98303214947c1": { + "parameters_hash": "b0f8496aea789eb43c3456efa898dec0", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:37:06.144215", + "end_time_simulation": 9.979999542236328 + }, + "5c35a1ea45a15002911d909dfb485adf": { + "parameters_hash": "2c1fd59f18317556cf021af499a3e365", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:37:08.372228", + "end_time_simulation": 9.979999542236328 + }, + "5cff394d93ba45fc950eb525d975c078": { + "parameters_hash": "a558ed4f0e1075ea1aad7f0f57281e41", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:46:05.519261", + "end_time_simulation": 9.979999542236328 + }, + "12ca7ab989435bd0860649a74648e1a8": { + "parameters_hash": "983e7c5bc271df9f2f739aa9bc13394d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:08.835941", + "end_time_simulation": 9.979999542236328 + }, + "b9f73c1ba2e14456bc1cb2701b0d61d4": { + "parameters_hash": "cf99288aa0ffe147b58431d2f616b3a9", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:46:11.832091", + "end_time_simulation": 9.979999542236328 + }, + "509ca665d1854fa596c901a5468197ab": { + "parameters_hash": "edf4a264ac6a522c53e7ebd3aec2e085", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:46:15.221987", + "end_time_simulation": 9.979999542236328 + }, + "c76e0f449bfa49c78d9a85af8aa2dc97": { + "parameters_hash": "afc610b7b01f7226932ec3895de7c6b3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:18.305634", + "end_time_simulation": 9.979999542236328 + }, + "d913f421c0964aea99408cec015c64c1": { + "parameters_hash": "5e68d80623386912e06fcdb7dd6ea844", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:46:21.881940", + "end_time_simulation": 9.979999542236328 + }, + "745b3507edf25938bb0e60b6f80a15b0": { + "parameters_hash": "e1c9d203168591e4b6cbaf8fdf005cd0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:25.561004", + "end_time_simulation": 9.979999542236328 + }, + "8ffce32ebf95402db33ab77ea1fe5f12": { + "parameters_hash": "5770963896b1f72c958fafb47d6b7233", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:46:28.317395", + "end_time_simulation": 9.979999542236328 + }, + "5e690461565344e6a316bfb32c73c8bc": { + "parameters_hash": "dd864f3d6b1abdca9b95b0c469caa657", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:31.658755", + "end_time_simulation": 9.979999542236328 + }, + "421875a8a7f441c4bffd546a137058b9": { + "parameters_hash": "1d9f83b3539e5a79fade52f3c72e9ab0", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:46:34.717972", + "end_time_simulation": 9.979999542236328 + }, + "cf7f5181b77845a8ae91754abcfc5bc5": { + "parameters_hash": "736e3925f27e3dddf8f8c62382e2e1d2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:36.994723", + "end_time_simulation": 9.979999542236328 + }, + "c320a1e1a03d448bbdec9cf368747927": { + "parameters_hash": "8ee91c3b869b1aebe1bf2bb4bafb9197", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:46:39.517918", + "end_time_simulation": 9.979999542236328 + }, + "2992f1216d8143aeb1d36a73e17f8f60": { + "parameters_hash": "94f521c2c8b8996d178688c4cf6efdc0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:42.205054", + "end_time_simulation": 9.979999542236328 + }, + "e14b900a3869432ba60969deadf8685e": { + "parameters_hash": "84c3b748a42c9e645419b159855238a1", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:46:44.962138", + "end_time_simulation": 9.979999542236328 + }, + "bf6cd10c66ec5575baaaabbbe51c34f8": { + "parameters_hash": "211e24afc2c3505061d6a7b6d882ff20", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:47.233093", + "end_time_simulation": 9.979999542236328 + }, + "5610326a6f3c4deb963f3a6e680ad816": { + "parameters_hash": "25d06f68fcc58c07c32ffd700c6b87b9", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:21:32.967008", + "end_time_simulation": 9.979999542236328 + }, + "275f9d36f018502eb8e901d157938d8b": { + "parameters_hash": "e642d1140c057a95a560a1c05d462c47", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:21:35.065633", + "end_time_simulation": 9.979999542236328 + }, + "afa4989867684534869139ce3935758e": { + "parameters_hash": "d2d6c6fc410a5daaf1a7e8d9810639de", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:37:11.138994", + "end_time_simulation": 9.979999542236328 + }, + "e51374bc16115166aa3c83bca453efa9": { + "parameters_hash": "2f9cc86360b4abd253df2f52dcbf950d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:37:13.769113", + "end_time_simulation": 9.979999542236328 + }, + "e2f5821c6d6d4efcb5885f28802556fa": { + "parameters_hash": "b6649eedd523532c7370c7dffcb1ac29", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:46:50.582585", + "end_time_simulation": 9.979999542236328 + }, + "f04bf5c8dd6e55e3831cb38d4d927d50": { + "parameters_hash": "a826eee6e0b89fc95bdf4baeaae914b0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:46:52.947263", + "end_time_simulation": 9.979999542236328 + }, + "52cb93e7e69d4abf85d6db293440f76c": { + "parameters_hash": "a9cc86f19309fc05fdef3a4fc2de77d6", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:46:55.454349", + "end_time_simulation": 9.979999542236328 + }, + "7a3228b4dcf24621821132073af92aaa": { + "parameters_hash": "86b90eca814b3fcc43c81601070bead7", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:46:58.570815", + "end_time_simulation": 9.979999542236328 + }, + "a10344525aae4f02ad2de434f897f0cb": { + "parameters_hash": "dce6f654409bf1ecd3c665417576c260", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:01.404394", + "end_time_simulation": 9.979999542236328 + }, + "10aa7e13ee764f2084f91a6b31c9fe06": { + "parameters_hash": "ded481d17742912a414e89bb6bde0bfd", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:47:04.413880", + "end_time_simulation": 9.979999542236328 + }, + "928f92f647225ff786710bb6af7349e8": { + "parameters_hash": "8e2e9032ef0560b0c9c8e28d799a5fc1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:07.238298", + "end_time_simulation": 9.979999542236328 + }, + "a37e9dbdfc5c4bcfa2da4fb4539dba82": { + "parameters_hash": "379a9f025d5734830605d30c8561d0c2", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:47:10.014721", + "end_time_simulation": 9.979999542236328 + }, + "4f9c0d24a44542589683ecb4d9c05310": { + "parameters_hash": "095679447fb9e86a0abd48c1980f4b2d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:12.161887", + "end_time_simulation": 9.979999542236328 + }, + "f664534ad3bb423eae4c699ad1d1ad79": { + "parameters_hash": "7ec41326259c36505116a76220602663", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:47:15.738651", + "end_time_simulation": 9.979999542236328 + }, + "83caaea402834b4d97147b9beac4aa6c": { + "parameters_hash": "6e130745e040fde7b28db59fb674cd10", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:18.594770", + "end_time_simulation": 9.979999542236328 + }, + "85ae5ed308024390843247516b5920a3": { + "parameters_hash": "57c447f10c55c5b2c1880276bd062174", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:47:21.462507", + "end_time_simulation": 9.979999542236328 + }, + "7ceff1c2962e492b95eb3acd1a80438c": { + "parameters_hash": "b932eaf587380966303bb1167f6070cb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:23.980812", + "end_time_simulation": 9.979999542236328 + }, + "292da126e2c945548c230bd6e86615c5": { + "parameters_hash": "3afb454d1fb0d2c073d7ffb0514524b0", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:47:26.677983", + "end_time_simulation": 9.979999542236328 + }, + "366f5e3753e4410b8dcd3517201d3c84": { + "parameters_hash": "0d7844f50058e0e32402f39373d166d0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:28.883842", + "end_time_simulation": 9.979999542236328 + }, + "a2464a5357c44faf9c7426be7efbf8b5": { + "parameters_hash": "e297a9c4f44c860f61691542211ade9b", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:47:31.454849", + "end_time_simulation": 9.979999542236328 + }, + "4e6967785ed04078a2bb35c9fee08d57": { + "parameters_hash": "34b47ac6e2154c3c8fcd790843429057", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:33.732215", + "end_time_simulation": 9.979999542236328 + }, + "d55fca9c48a04f4498d553c8b58d37a0": { + "parameters_hash": "969036abdd7d4ba272b2e932dbf1879b", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:37:17.292800", + "end_time_simulation": 9.979999542236328 + }, + "421a65fe13385826a68195bd2068785e": { + "parameters_hash": "c2951a2c135df722cf25839463c5476d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:37:20.006827", + "end_time_simulation": 9.979999542236328 + }, + "a3dc646b3f7647079bfb49c9acaa18e6": { + "parameters_hash": "abdfb62459ed77149a2e3dff9e494c72", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:47:36.465318", + "end_time_simulation": 9.979999542236328 + }, + "a924ecd1852456539eb5e3ed509ecc09": { + "parameters_hash": "ebfce3b72f3bb24c6c9ad1a5f40b817c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:47:38.591793", + "end_time_simulation": 9.979999542236328 + }, + "a8e9bd401ebe4e9fb6d22cb8c5acdb47": { + "parameters_hash": "5bfcf479890a884a29b280dfda9da316", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T16:47:31.502510", + "end_time_simulation": 9.979999542236328 + }, + "6ee97ea8a1264811b558b1c747ca0188": { + "parameters_hash": "f67f44b42964ac83d85eda8b73940bf4", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T16:47:37.147084", + "end_time_simulation": 9.979999542236328 + }, + "8d580d57423049e28e71d02c65ff9626": { + "parameters_hash": "01aae10e825804183ea801c0986131cc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:47:40.403573", + "end_time_simulation": 9.979999542236328 + }, + "00af86611bf0452986e018425db6cdd2": { + "parameters_hash": "0e4b8232db2bfc1db623f7fe255a9e36", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T16:47:43.678970", + "end_time_simulation": 9.979999542236328 + }, + "d8690d006f3755e48386eaac9ebe31ec": { + "parameters_hash": "ef8db85b8e16dc45ebde399185321dab", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:47:46.222275", + "end_time_simulation": 9.979999542236328 + }, + "4e7013b8be8b4db2b761dd6a135aa98c": { + "parameters_hash": "b1c1f900eadf69c8765360c3779f270e", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T16:47:49.059164", + "end_time_simulation": 9.979999542236328 + }, + "11412a2988774df09ca6cdbca7292d6e": { + "parameters_hash": "1f4ed8eaf32079e52a01b64ab7899045", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:47:51.687923", + "end_time_simulation": 9.979999542236328 + }, + "abda5fe0cf46427fb355f3177ea223d9": { + "parameters_hash": "6219258672359c708a279a9173f56b52", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T16:47:54.558567", + "end_time_simulation": 9.979999542236328 + }, + "ce129c08df6a48c4860db63123905f26": { + "parameters_hash": "566a74047cf7010ddf8e84c1f6f6540d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:47:56.784320", + "end_time_simulation": 9.979999542236328 + }, + "ae1f65c141714e29a48a2b202e47050b": { + "parameters_hash": "888bee0a555777a9a62ae1abf39383d2", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T16:47:59.375682", + "end_time_simulation": 9.979999542236328 + }, + "6f9506608c874f98aa9b36f9f2fad551": { + "parameters_hash": "e1feb89c3e6c1ab34bf218743da8ae7a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:48:01.795783", + "end_time_simulation": 9.979999542236328 + }, + "499ee4d0d6704a5488c592033341bb51": { + "parameters_hash": "65c101cabfc333d4c36902b29aa6d952", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "failed", + "timestamp": "2026-05-05T19:22:12.836696", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "a323bcead7315cc8b07b5ffaf4378447": { + "parameters_hash": "ad98823ef3744c66a51ace4e1d6e8c31", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:22:14.235869", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n" + }, + "477fa4f8ca014f559465bdc2c2318c78": { + "parameters_hash": "7b1c46eae667792fb1c46e960dc99993", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:48:04.507053", + "end_time_simulation": 9.979999542236328 + }, + "96f01eb7fbf24e3f87251c5598a18f81": { + "parameters_hash": "3bffa870dc533a6afff1993c5542e9e1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:48:07.127767", + "end_time_simulation": 9.979999542236328 + }, + "b9d5d04e374c47d58d820fca9d364a11": { + "parameters_hash": "98b81617fc2dfa7a56b351b51a3c4c22", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:59:55.634809", + "end_time_simulation": 9.979999542236328 + }, + "bffffee1d39255e3a6e19646379ac1fd": { + "parameters_hash": "ab78fa60ecbca913f94b2bb86a2baa38", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:59:58.699361", + "end_time_simulation": 9.979999542236328 + }, + "15aedbc1d25c48a8a238da20d7dd6f9c": { + "parameters_hash": "da953b5f7c8eb6b68182865041864921", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-05T19:22:17.266055", + "end_time_simulation": 9.98 + }, + "3cfb0baaca795fec8cbf3bfaf3fc1e06": { + "parameters_hash": "8379c4ed287dfd8eb22235e52027f893", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:22:19.925858", + "end_time_simulation": 9.98 + }, + "cc258dc235904463b1dba49ca7c1b97f": { + "parameters_hash": "27f4904e5eb143dbc07599387231640c", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:00:02.047357", + "end_time_simulation": 9.979999542236328 + }, + "c83ac21e681c4832b44640577489c6bc": { + "parameters_hash": "3d77bcff006459a8b99cb9abfb6f0386", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:00:04.791519", + "end_time_simulation": 9.979999542236328 + }, + "3dec8274ede5430e93ead4b75fefffb1": { + "parameters_hash": "eb95aa8d127af109a83e2dceffcbc282", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:07.654501", + "end_time_simulation": 9.979999542236328 + }, + "66b4a8fdb8c84c82886aabde4b5f9a08": { + "parameters_hash": "1be94b3e6be263cccc46077fdd7ebfac", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:00:10.911263", + "end_time_simulation": 9.979999542236328 + }, + "9138da4cf2d854e1a9023771dd0be754": { + "parameters_hash": "cb39526fe9bc9de5e4492689f38cab61", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:13.436009", + "end_time_simulation": 9.979999542236328 + }, + "baf1e1790569451993d611a8cb903120": { + "parameters_hash": "eacc4946d6f5b022256b47ba95bfc884", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:00:16.330417", + "end_time_simulation": 9.979999542236328 + }, + "d7736ed52dbc58079432937fd2a06943": { + "parameters_hash": "56677c8593d95c7a9695892223c6e582", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:18.977191", + "end_time_simulation": 9.979999542236328 + }, + "bd084d19f2444af8a6f98ec7815d2882": { + "parameters_hash": "5aef2efd606ea97c48461efac8cfe20e", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:00:22.387015", + "end_time_simulation": 9.979999542236328 + }, + "28df53dc4ee54a6d816e512f84c02e18": { + "parameters_hash": "c8c47333c26d7595525cb21607d3627e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:25.204314", + "end_time_simulation": 9.979999542236328 + }, + "06557b2cc73a4f1e8d7662f6931a630c": { + "parameters_hash": "beb402f27a3a4898e071dbc29aaeb8b8", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:00:28.383750", + "end_time_simulation": 9.979999542236328 + }, + "d12a555af94f4b169c7a0eb13e067a62": { + "parameters_hash": "f0bdfad6111cf101dfb6dd76f0c74aa4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:30.846099", + "end_time_simulation": 9.979999542236328 + }, + "c8c7a5a6c6cd4176a078d38c7663c01b": { + "parameters_hash": "5cade7432107a0f3f42e3bc283ecfe25", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-05T19:22:22.453059", + "end_time_simulation": 9.98 + }, + "d8b77f327ec251dba04babcfd96c3678": { + "parameters_hash": "a0092f972f0751ec402ff316093fa6f5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-05T19:22:24.570087", + "end_time_simulation": 9.98 + }, + "89120e47aa19439992f66c03bf231af3": { + "parameters_hash": "e30cdadbf3dea837ce7927d5a48041ae", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:00:33.539575", + "end_time_simulation": 9.979999542236328 + }, + "8d59e9561cb95164a587bf58bb4b8fd4": { + "parameters_hash": "8a83213ad00c73056ea766ba6d59bd87", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:35.873501", + "end_time_simulation": 9.979999542236328 + }, + "59d90dcd97cb4b84ba9d4bc95bdf2985": { + "parameters_hash": "e62db7cbe306be7fbc470a4d2462fc96", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:00:38.332586", + "end_time_simulation": 9.979999542236328 + }, + "24dc598341aa4674afba02f25e262ce5": { + "parameters_hash": "7656301d2e0b15ab42441a38c0c482b3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:40.625055", + "end_time_simulation": 9.979999542236328 + }, + "d2db848c867a4ffd9b6bc42d6df217d7": { + "parameters_hash": "33e9e6ab68ae56eae253c4f9412111cc", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "failed", + "timestamp": "2026-05-05T19:22:37.104868", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "a6dd49d400ce4a328d65184a0375930f": { + "parameters_hash": "831d973370f81a96f4218ab16a11c14e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:22:38.185820", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n" + }, + "aec60d90f7ef4979a80c56f5c5bdeafa": { + "parameters_hash": "7dfdef65098f35fb6f5804a59777d9cf", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:49:08.654932", + "end_time_simulation": 9.979999542236328 + }, + "1e7a9a8f686f49d38a1791a7d96016de": { + "parameters_hash": "c4860a63006f2e64518b5f10795c9ed2", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:49:11.991730", + "end_time_simulation": 9.979999542236328 + }, + "9ab11e47426f4a8280d4101875ccb04a": { + "parameters_hash": "583dba97f1f7b321d5506f779944f47a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:14.430420", + "end_time_simulation": 9.979999542236328 + }, + "e4a4fea8b64e496cb7e24f4de9f270b7": { + "parameters_hash": "f4928cc375bb9b021a302e566b6c3062", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:49:17.684705", + "end_time_simulation": 9.979999542236328 + }, + "4d19c44774c45d67bbe3ca0a4bf15809": { + "parameters_hash": "32289612bc66ddb952de3a3cec90d1fb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:21.003204", + "end_time_simulation": 9.979999542236328 + }, + "23d9613659ea4258910b9c48eb7cf01f": { + "parameters_hash": "70766190301f52668303747b5d6d8d76", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:49:24.074093", + "end_time_simulation": 9.979999542236328 + }, + "f2dcb5feeb7b5ffd84bd448ff463f9e1": { + "parameters_hash": "e0e6ce595730daacc64c322d2b3431ac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:27.179629", + "end_time_simulation": 9.979999542236328 + }, + "08134d4da4b64c78a881feac607a2e5b": { + "parameters_hash": "7be370c6e6cd840b1ca2e305fae22c32", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:49:30.063423", + "end_time_simulation": 9.979999542236328 + }, + "869af35a00cb46f9887ee2f015ba226f": { + "parameters_hash": "c5ca24ce35054fe227e2fdbcb69fabf1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:33.212458", + "end_time_simulation": 9.979999542236328 + }, + "50d92d7b10f64b2da1de593bed3146d7": { + "parameters_hash": "124546f3141f420dd0f0b32dba9f42f1", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:49:36.079560", + "end_time_simulation": 9.979999542236328 + }, + "a8ab65041c0044a8a89622d2e80e01bb": { + "parameters_hash": "85f1de77874da742f379cff1dc575366", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:39.193124", + "end_time_simulation": 9.979999542236328 + }, + "e6e67852460340c291d3b2d33978e955": { + "parameters_hash": "988c36d05934128c74294961821599ef", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:49:41.956329", + "end_time_simulation": 9.979999542236328 + }, + "0c8146862288559f95205b971459dc7a": { + "parameters_hash": "0d848361a874f5a37403f8f4a8fb1207", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:44.638128", + "end_time_simulation": 9.979999542236328 + }, + "dcc54ad4f3014c0cb2545e1f84a9ee38": { + "parameters_hash": "cffa3fdf64e42d8be45cfebcbd63ee46", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:49:47.533046", + "end_time_simulation": 9.979999542236328 + }, + "1e053ae04ad94ec28a748447e73689cd": { + "parameters_hash": "c25d76f93707e10da6b5f7e9f6ec98bd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:50.300657", + "end_time_simulation": 9.979999542236328 + }, + "1a2cf72777984203b112ba42778278f1": { + "parameters_hash": "6e309b0cb4e2be35bb8b4bce5d5ce6fe", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:49:52.955066", + "end_time_simulation": 9.979999542236328 + }, + "6a9c143732694a2fb52bf39ef414de74": { + "parameters_hash": "f0451b878079db9080daba9953850b77", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:49:55.941711", + "end_time_simulation": 9.979999542236328 + }, + "9e2271b4625541f7964a03844c658511": { + "parameters_hash": "b8827508f18d571c18ad4345357f8c5f", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:49:59.123528", + "end_time_simulation": 9.979999542236328 + }, + "c4b27d78e2c9592c86f04555c12273c1": { + "parameters_hash": "a0cf8b1daf4d1f9a9a1681de390f6a35", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:01.816738", + "end_time_simulation": 9.979999542236328 + }, + "c56def5d135e40c290812174a4d3a066": { + "parameters_hash": "4cf0225133f9691c3e863ab18a5f2bb3", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:50:05.473076", + "end_time_simulation": 9.979999542236328 + }, + "e927c665f3f5465c8a53145bae0dfd36": { + "parameters_hash": "41df09e06e0fe8b8ef774a713510ec32", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:50:08.440772", + "end_time_simulation": 9.979999542236328 + }, + "79d6b156a2b0470bb17772938b9c4da3": { + "parameters_hash": "457be17d2f2939a59179c741c382ff18", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:11.470707", + "end_time_simulation": 9.979999542236328 + }, + "01270c870ac94f1d8ec11e9fd23f327d": { + "parameters_hash": "76ff5a9598f60dd3d05ccf716e6500ea", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:50:15.541476", + "end_time_simulation": 9.979999542236328 + }, + "ce26d085727e506592df57c37c36fb75": { + "parameters_hash": "b2dcb7ad324ae16e3753541bcfca5098", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:18.538187", + "end_time_simulation": 9.979999542236328 + }, + "cfcf113e827240a6b7e7a4be86078864": { + "parameters_hash": "c96b86d4ccfaacdef36a59b38587aae5", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:50:21.417780", + "end_time_simulation": 9.979999542236328 + }, + "be7a5990092d58d19aae77403ef84f4e": { + "parameters_hash": "c9b20542fa4b3c88b7b821b389f31ab6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:23.820016", + "end_time_simulation": 9.979999542236328 + }, + "95357a192b404d9d942b699a96f3f91d": { + "parameters_hash": "9f0e10f22186a49d6e156a0074913a6c", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:50:27.017440", + "end_time_simulation": 9.979999542236328 + }, + "40cae27f4ade43fb8ea906cfb1cc6421": { + "parameters_hash": "6505a27f02b170977c89fda72c5241c3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:29.627827", + "end_time_simulation": 9.979999542236328 + }, + "01ca63e4fb7d4648a1ced30b6a4ecfba": { + "parameters_hash": "afe46de9d1743013a555124c7bd443e8", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:50:32.179533", + "end_time_simulation": 9.979999542236328 + }, + "2485cdff05454c7483aa4a992a6ed1f6": { + "parameters_hash": "a78a832e092eec3455caabb8d16cbb76", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:35.246584", + "end_time_simulation": 9.979999542236328 + }, + "0faf91cbf206423294c8646f226f9b9e": { + "parameters_hash": "adfd8ba9186266b08be14a022b76a10c", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "failed", + "timestamp": "2026-05-05T19:22:50.301912", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "25c12ad712b159f9a0625e7ae851b8a4": { + "parameters_hash": "7ec9dca1c05f334cc7a72aec0a20b705", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:22:51.514628", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n" + }, + "006980367df84d448f547ac755d09e28": { + "parameters_hash": "6948a5fdd18600274409763e5c3f0d1f", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:50:38.462468", + "end_time_simulation": 9.979999542236328 + }, + "4a405d3b5f2940b2b1f00ddf84c7aa87": { + "parameters_hash": "e3990eb741eec52e46b54deefe1cc09a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:41.230388", + "end_time_simulation": 9.979999542236328 + }, + "743ece0da32443a68b23b49ae020b720": { + "parameters_hash": "55ae5fe8756330e587f987abe3c136c9", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:37:28.718040", + "end_time_simulation": 9.979999542236328 + }, + "ceb6f5dbb3d3516184608b50ae18c4bc": { + "parameters_hash": "5ce2b0d0b4afa9a30399e0c6a2677c70", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:37:31.314207", + "end_time_simulation": 9.979999542236328 + }, + "8d76ad585ba14e95a142ac12e1150795": { + "parameters_hash": "b968628a61ab2d453e577c8591da2bea", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:50:44.604413", + "end_time_simulation": 9.979999542236328 + }, + "f33d59af115355839fd1b89a343f4754": { + "parameters_hash": "230a6afc33bcde87c7937660a6f0c4b0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:47.621464", + "end_time_simulation": 9.979999542236328 + }, + "9f2a4f73dc22406a8e8277f1a83ed265": { + "parameters_hash": "0648ed0cc0dc962d7c29e8ad7495e57e", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:50:50.688932", + "end_time_simulation": 9.979999542236328 + }, + "a7174cc25d9744eab005ec3b4a45fb2d": { + "parameters_hash": "28d269f6fce7b064f81b37fd5e0746d5", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:50:54.286734", + "end_time_simulation": 9.979999542236328 + }, + "76f6ea63fd9f43c3a0d3cf0f588cfd01": { + "parameters_hash": "205f1a855157ec98e696875fc0fac5b2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:50:57.080116", + "end_time_simulation": 9.979999542236328 + }, + "38e9b9cde3d547a9924b3af04070203a": { + "parameters_hash": "eea58208b52bebf71aa8d5e56c4b0b07", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:51:00.538929", + "end_time_simulation": 9.979999542236328 + }, + "3117d3709f0c5bc4bcb9a8bca1347973": { + "parameters_hash": "1790c34684925c8fb5679819f198680c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:03.324695", + "end_time_simulation": 9.979999542236328 + }, + "b746721382cf4b5bb7a678c9d5fd9d86": { + "parameters_hash": "d69aed2208f9ba14e356c03bae41eb04", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:51:06.632420", + "end_time_simulation": 9.979999542236328 + }, + "ae11b355d4185e37bc29b5ff0e50a1ca": { + "parameters_hash": "ccdf0d6fe31746544432cc22977374ae", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:09.717399", + "end_time_simulation": 9.979999542236328 + }, + "6ff497703dc54edb8534f212c5b7c567": { + "parameters_hash": "b4c9d0e281a2bf124b7908e603c77505", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:51:13.007890", + "end_time_simulation": 9.979999542236328 + }, + "1aaa0186dd124311a84f1f6d4f0ec8ac": { + "parameters_hash": "2e7577c5d677a9b079e38d4716edb435", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:15.595805", + "end_time_simulation": 9.979999542236328 + }, + "d0459a5a08cf463493c8e85298a47421": { + "parameters_hash": "c1592f39fe8d25c39601df14244963ab", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:51:18.530132", + "end_time_simulation": 9.979999542236328 + }, + "eb106d27a2bf4dd4acc894a274e7f231": { + "parameters_hash": "82c2f746478d6a55d2cad421e9e5036c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:20.692757", + "end_time_simulation": 9.979999542236328 + }, + "458fa65a9d19430ea2c93b693e98523e": { + "parameters_hash": "85a9268b22b68c1cc8374a101593e941", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:51:24.004687", + "end_time_simulation": 9.979999542236328 + }, + "c4f31dedf90b4ea4b6056ab690aabbec": { + "parameters_hash": "f5b8f787693fcdde8b4cdbbd68e414ee", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:26.306965", + "end_time_simulation": 9.979999542236328 + }, + "e6f0f2340188474fabe7182a920d967a": { + "parameters_hash": "4505e41f50eecfff868f4f1ed272d7cc", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:51:29.721911", + "end_time_simulation": 9.979999542236328 + }, + "31f3971395ed5d448b5cb65a5bfab840": { + "parameters_hash": "594a15f821888f6a9adf50ac4e68fab4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:32.701778", + "end_time_simulation": 9.979999542236328 + }, + "55e6f899ce59474fba6e2f6707f830d3": { + "parameters_hash": "b8f79177370a9da7b10487b055c574de", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:22:36.193745", + "end_time_simulation": 9.979999542236328 + }, + "6b8ad51b45275017bb6b03544c05ce3b": { + "parameters_hash": "ad84188dd224c05f9782ca4f10e5be1c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:22:38.797483", + "end_time_simulation": 9.979999542236328 + }, + "e9265cb5151a4147be23993b13a8af42": { + "parameters_hash": "97093ad3a7d40c0f673de9e24211be91", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:37:34.091734", + "end_time_simulation": 9.979999542236328 + }, + "fffc272d3dc258e1b365a96c71e06fc9": { + "parameters_hash": "062cd2776b72519d22fbfd6dc87dd166", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:37:36.756925", + "end_time_simulation": 9.979999542236328 + }, + "633f8b94b4a64afab5c4183b67987092": { + "parameters_hash": "d3bacc516c99297eb00354bca60ba401", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:51:35.472939", + "end_time_simulation": 9.979999542236328 + }, + "f77f598d091b5f62ae6d05f69a63ada8": { + "parameters_hash": "784805862c1f6f3374aa70144834ec64", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:38.576630", + "end_time_simulation": 9.979999542236328 + }, + "472873131567464dab68c8060f142ec0": { + "parameters_hash": "cfaa188f1ed58f9ee3693fa453df2c38", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:51:41.321031", + "end_time_simulation": 9.979999542236328 + }, + "3ba3a08aac1441ab9f7062076ab2e050": { + "parameters_hash": "e5ba8922218dd2839d29d948257459dd", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:51:44.056264", + "end_time_simulation": 9.979999542236328 + }, + "441dc419b2bc433b99c7bb9ae4e244d4": { + "parameters_hash": "c3fb9c6bffcc7f6122b7e42738475c74", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:47.028529", + "end_time_simulation": 9.979999542236328 + }, + "c77ab99ba98b424f84e63a4a382fb955": { + "parameters_hash": "c1867be37bfbce4d5850c40a9390dc2e", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:51:49.900926", + "end_time_simulation": 9.979999542236328 + }, + "efa51a3a4283589195d81c7eaaa80b50": { + "parameters_hash": "2fe50c5e973eab642b37ea1dd1b6bd62", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:52.432607", + "end_time_simulation": 9.979999542236328 + }, + "3c1ece8357064e529486c63fc775577a": { + "parameters_hash": "216b099f3d28f4ebc7606c3b9f5d27a8", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:51:55.878278", + "end_time_simulation": 9.979999542236328 + }, + "d08ea8c09ec25749a659e09bf4975c3c": { + "parameters_hash": "c2d31c72039b4ed9a05de16eedc5de9f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:51:59.015177", + "end_time_simulation": 9.979999542236328 + }, + "2eb3d48282394e63b5d81f7c900525c8": { + "parameters_hash": "1d4c1b60445675d16a6d85dffb6f86e2", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:52:02.437660", + "end_time_simulation": 9.979999542236328 + }, + "2d052d08ab1d42179e1726a9db510054": { + "parameters_hash": "849aab112317e29912acacccf91cc405", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:05.299804", + "end_time_simulation": 9.979999542236328 + }, + "2431ebd6801e4e988b2f218a970bf931": { + "parameters_hash": "96bfc5ef229587d96678a366027ed537", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:52:08.711886", + "end_time_simulation": 9.979999542236328 + }, + "2cac78149e9544398425ec60e798e0fa": { + "parameters_hash": "b01f60a85c2d53105de6fe2ffea10972", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:11.299357", + "end_time_simulation": 9.979999542236328 + }, + "783e79410558471f89a231cd01a289d2": { + "parameters_hash": "e42f0778316bfe0d2e2f04dbba1df92b", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:52:14.517392", + "end_time_simulation": 9.979999542236328 + }, + "a64e4df98d754b95b485985bfb104e31": { + "parameters_hash": "412b62051eb4c1f22b14c41e8af5ba5f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:17.639712", + "end_time_simulation": 9.979999542236328 + }, + "bd222ab0a4034903923ada8b7a09816f": { + "parameters_hash": "423f2fec735b03ca2a4f38628bade3a8", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:52:21.170349", + "end_time_simulation": 9.979999542236328 + }, + "75c064a276e74606a0fa759419228450": { + "parameters_hash": "d005a66c0294a36c3e0c1bd942471849", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:24.130309", + "end_time_simulation": 9.979999542236328 + }, + "28aa7cd7267a4ef3be66ae0819874cbe": { + "parameters_hash": "38fa981191567abd07f351e99076475d", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:52:27.209481", + "end_time_simulation": 9.979999542236328 + }, + "3df9ff20a5714a1984f7e3ce324a6231": { + "parameters_hash": "286a247cb5b450cf97da8ec8f05990ab", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:30.009410", + "end_time_simulation": 9.979999542236328 + }, + "b1fb837092144b5cae1c2a952a2ea10c": { + "parameters_hash": "e6161e86609d22c3e2b4a980d1aee886", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T19:37:39.942561", + "end_time_simulation": 9.979999542236328 + }, + "9e5cf7287ba55f5db8bf304cc5811d59": { + "parameters_hash": "54e2ce1888fec7f846017c01eed93b82", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T19:37:42.314174", + "end_time_simulation": 9.979999542236328 + }, + "1b279f64d97b4ea8b68cf572fe829122": { + "parameters_hash": "bb7e757ade3c89977a820f10a09b7b89", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:52:32.867392", + "end_time_simulation": 9.979999542236328 + }, + "5eea7258f1555848acfc893d5a6bb12a": { + "parameters_hash": "1cbf7e58e0e7e5cbea05152f7e811a04", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:35.508064", + "end_time_simulation": 9.979999542236328 + }, + "e07a5edf78e844b8ba42f7b4add2cfb7": { + "parameters_hash": "48f61cda56b12176da48acee13424ecb", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:52:38.210446", + "end_time_simulation": 9.979999542236328 + }, + "0982071fbb684307913a380272a1ae91": { + "parameters_hash": "b2fd2ce49c308dfdd3b0ab309fa8e33c", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:52:40.580768", + "end_time_simulation": 9.979999542236328 + }, + "2c278400a9d443568c2940c5624cd2bb": { + "parameters_hash": "d233e4a9a57534c4738f720550159866", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:43.084793", + "end_time_simulation": 9.979999542236328 + }, + "af1e5fe87b2e4cc3a25b933ddb4b8841": { + "parameters_hash": "d84559841816455250f96df60f33eac9", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:52:46.660505", + "end_time_simulation": 9.979999542236328 + }, + "d04150859ad653fcbcf515e93691e443": { + "parameters_hash": "84489e906bfc93f222031bb096cefb26", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:49.609915", + "end_time_simulation": 9.979999542236328 + }, + "e8e9190868644fc38aa1113756e04fee": { + "parameters_hash": "5a5f4d277150617f1f739a6e55cc29e1", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:52:53.013909", + "end_time_simulation": 9.979999542236328 + }, + "dc904e57b7df59529a248ccb3e7cc459": { + "parameters_hash": "9a7dc7a7aed5a7d2a2d59e5607995953", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:52:55.670573", + "end_time_simulation": 9.979999542236328 + }, + "125c9571ba0845fba3df8d68a2a14fad": { + "parameters_hash": "4fa78bc2224ee15d36dcdf44de93454c", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:52:59.173273", + "end_time_simulation": 9.979999542236328 + }, + "c93d2783e21744c1aebd2d3a18847208": { + "parameters_hash": "2ede1e91a4f5846dbcba15c5f95e2030", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:53:02.162605", + "end_time_simulation": 9.979999542236328 + }, + "ad8867dbe72a470a98b2182b05a43c5f": { + "parameters_hash": "6fac1f29db17ed5ad31ff3b23cc3db87", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:53:05.063176", + "end_time_simulation": 9.979999542236328 + }, + "b9cb3bf45b10424ea82cf812767ec9f0": { + "parameters_hash": "1d7e5bd1440219fcd8178f087beae17b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:53:08.028914", + "end_time_simulation": 9.979999542236328 + }, + "7d55aa38010f47ca9275729033167bdd": { + "parameters_hash": "d01b80d1457380ac7f6fddc0f087ccef", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:53:11.240222", + "end_time_simulation": 9.979999542236328 + }, + "7b9797212bff4fb8858aaecd1187559d": { + "parameters_hash": "1abb432b6b9158fda1e2afcada443e67", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:53:14.877627", + "end_time_simulation": 9.979999542236328 + }, + "8bc590e7d4274e4db6c5ee5ef76a48d0": { + "parameters_hash": "3f979decd65737a4f9f13b0cd8a9ba9a", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "failed", + "timestamp": "2026-05-05T19:23:03.870560", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "92ed561f7a5653ea8567b7ea9093f23f": { + "parameters_hash": "e4cd499bcf6e8b0ab688b1da6f9e0b98", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:23:05.007234", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n" + }, + "2aca128ddd88437dba2a531e84298987": { + "parameters_hash": "9989f76334c8bb91f14e7792d75fc88a", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:53:17.921720", + "end_time_simulation": 9.979999542236328 + }, + "5226a920492c4b3b99061efd112ecdda": { + "parameters_hash": "f16a563fae74a2882d707ab8506ab8e6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:53:20.687174", + "end_time_simulation": 9.979999542236328 + }, + "0c5a280f084b4d1fa52dddd30269c1d4": { + "parameters_hash": "8f3252478336ac13317e41fe351d861d", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T16:54:54.926548", + "end_time_simulation": 9.979999542236328 + }, + "92277f4a3b0056f1b7be37cb0ab9941e": { + "parameters_hash": "ce6ecb42253bc9f89e85c588863cf804", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T16:54:58.664563", + "end_time_simulation": 9.979999542236328 + }, + "95b6c252820d4c82805315cb6f80d529": { + "parameters_hash": "9b365474216d2b776e9596272d3c3661", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T15:50:17.200690", + "end_time_simulation": 9.979999542236328 + }, + "6d1ccd9c58fa50f98d40f94f7112b732": { + "parameters_hash": "30f1cc16c97ab0bc2500caf88c26e844", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T15:50:19.941637", + "end_time_simulation": 9.979999542236328 + }, + "38482b8747bc4bfe88ac617356e9c6fe": { + "parameters_hash": "1bf32ba6559ba683e01da61006480882", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:00:43.430363", + "end_time_simulation": 9.979999542236328 + }, + "ff3a19d5eaad46c1aed4ca4373c26de3": { + "parameters_hash": "f27adfcf89db0298da43ee6f363db29e", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:00:47.681559", + "end_time_simulation": 9.979999542236328 + }, + "770ead8ac6a843fe8986d330a26db2b5": { + "parameters_hash": "89ea87183fb33ad3ad3358c98e1e70a4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:51.894715", + "end_time_simulation": 9.979999542236328 + }, + "cbace2e8b8ea4f57a52cddece9ec485f": { + "parameters_hash": "216e3a0f31f61ccbe368fa642f5d4d6c", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:00:56.634264", + "end_time_simulation": 9.979999542236328 + }, + "7334062131e45af38c7116576cd1f63f": { + "parameters_hash": "c220a8946675906d01b30b39f1ff7872", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:00:59.703106", + "end_time_simulation": 9.979999542236328 + }, + "777005d52fa04b3c9842c921bb3166fa": { + "parameters_hash": "447380f8ff818229d1181afb522e8a30", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:01:02.477990", + "end_time_simulation": 9.979999542236328 + }, + "372be8bc88da5b87af3b41e863f3394a": { + "parameters_hash": "7c120605e2787cb37287f2811c3717ab", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:01:05.011784", + "end_time_simulation": 9.979999542236328 + }, + "1f0dabc340594e22b6dd59a52dc6dce7": { + "parameters_hash": "0e47ccac48db2494b42284424bf65579", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:01:07.729765", + "end_time_simulation": 9.979999542236328 + }, + "14123b2df18b447c98697e4971ce7628": { + "parameters_hash": "429cdd09376702bf9313712a697d5efa", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:01:10.062319", + "end_time_simulation": 9.979999542236328 + }, + "2f9a0daa94e44373b13b61004cc6fed9": { + "parameters_hash": "9de0f78ff3a2243556cea88e62a1d99c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:01:13.434523", + "end_time_simulation": 9.979999542236328 + }, + "91986c45e6cc41619458f40bc123f763": { + "parameters_hash": "f5a804e1f287fdcb8232d58c4575664e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:01:16.174600", + "end_time_simulation": 9.979999542236328 + }, + "53d61728255b43158f043258bf085cb9": { + "parameters_hash": "87ffc1178bf3675e97c2b9c7268f7ae0", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:01:18.629790", + "end_time_simulation": 9.979999542236328 + }, + "6fbea7e530344ecbbf1bcc5936f56880": { + "parameters_hash": "94067b9d491ba3305e44f0ea03c2816d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:01:21.023198", + "end_time_simulation": 9.979999542236328 + }, + "5eeff6be20ae44fc8a8ea24f3e384b3d": { + "parameters_hash": "e06f12e2ebc7f3e47b084a7d7136992c", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:01:23.609846", + "end_time_simulation": 9.979999542236328 + }, + "7e6fdc651fe552429217b93d0daf12cc": { + "parameters_hash": "48c1a67f5b4826498fd45740ae676e72", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:01:25.999630", + "end_time_simulation": 9.979999542236328 + }, + "bbaf00b6bbb8412d83fee39a553bb077": { + "parameters_hash": "473865f91d0a242cc341b33d35e2cd81", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:01:28.679090", + "end_time_simulation": 9.979999542236328 + }, + "40ff036056f95c66b9302d99a364bd35": { + "parameters_hash": "b11080ba8a9aa7381bb21b05ea7f05fb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:01:31.037732", + "end_time_simulation": 9.979999542236328 + }, + "cd7e56402b9f413291032827a4eebdcb": { + "parameters_hash": "5f3a6c2178968eed9b860a1b3af0d462", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:01:33.872347", + "end_time_simulation": 9.979999542236328 + }, + "138d88901d5d4b1292e5e823d55a0aec": { + "parameters_hash": "1b89b823bbf983c1e63a5e2beb6df1dc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:01:36.330159", + "end_time_simulation": 9.979999542236328 + }, + "31dcf19a3b73436f9e844e7822ac52c3": { + "parameters_hash": "1516bb9e64ac9ef6eef5a00383c589ca", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-02T16:54:16.941649", + "end_time_simulation": 9.979999542236328 + }, + "fb2367ca63f6468bbb95f4ddbd0d99c9": { + "parameters_hash": "f0150020d3dbb1c4558dd1b87ef56135", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:54:19.850871", + "end_time_simulation": 9.979999542236328 + }, + "62c94c3fcfff420b8353b8bb338ea521": { + "parameters_hash": "76f34a824433365c58468cd2435fb362", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:54:22.404441", + "end_time_simulation": 9.979999542236328 + }, + "50e366ba325c4951aa63b3c1e5d9962d": { + "parameters_hash": "0c3931b985a9b00a4858e39e53d0ebfe", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-02T16:54:25.822420", + "end_time_simulation": 9.979999542236328 + }, + "d0e5f1c2451d51efa2c9f3a40aba48af": { + "parameters_hash": "931c6cf34f2ec9a7e698a6fd323361e5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:54:29.073847", + "end_time_simulation": 9.979999542236328 + }, + "f6cf0dd947e24c92a93c79908c289154": { + "parameters_hash": "006c38f6187362e3d1ae38c643e8dd56", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:54:31.684465", + "end_time_simulation": 9.979999542236328 + }, + "83f0496526cc57aca50ea192b41bcd52": { + "parameters_hash": "d9513e7c2bde4d8ee4d30898aa2c1d88", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:54:34.109344", + "end_time_simulation": 9.979999542236328 + }, + "8840791a44ca45f2b1e72ee59093cef4": { + "parameters_hash": "e1221f8c6f6d1ecc6bf568022ba0796d", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-02T16:54:36.852683", + "end_time_simulation": 9.979999542236328 + }, + "35e0a4a1185746a1961538a0c6fbf936": { + "parameters_hash": "6a69cd05540bdc51a5308782876ab716", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:54:39.438142", + "end_time_simulation": 9.979999542236328 + }, + "c67a5acb75934bdbbc947a779778d1c6": { + "parameters_hash": "8896c58a33956d65b70f35fbe19f355d", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-02T16:54:42.506973", + "end_time_simulation": 9.979999542236328 + }, + "e12c4eaab43844259b46892a1eca7d34": { + "parameters_hash": "41c10685b6a1d4d07ab38efbe67ca326", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:54:44.771347", + "end_time_simulation": 9.979999542236328 + }, + "1f0a940c521643948cd6a4a3f8c15067": { + "parameters_hash": "fcb9df154efc9f87f1877b165535cc23", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:54:48.131585", + "end_time_simulation": 9.979999542236328 + }, + "160bba135c795849b8e286ed0df63683": { + "parameters_hash": "da956391c70a4fe36d92caf292637cea", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:54:50.825086", + "end_time_simulation": 9.979999542236328 + }, + "df19def46e7e4d4b9fe0aac39535a598": { + "parameters_hash": "a4f51b976fd0fd1cfcc86f99bc1e0063", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:54:54.178736", + "end_time_simulation": 9.979999542236328 + }, + "8493a49263cc456aab7eb21a15d296cb": { + "parameters_hash": "be1a38cf5fb6eb2b891a3e7366357a5c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:54:57.089902", + "end_time_simulation": 9.979999542236328 + }, + "3ad87b80437b43fda95e551a2efc7b87": { + "parameters_hash": "6daa1de025af0c11ca69b85d58be04dc", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:54:59.931373", + "end_time_simulation": 9.979999542236328 + }, + "fbab5a4d463040edb7855533156f8a8c": { + "parameters_hash": "f0b8cb3f5c0a7903a4ae48406922dbd6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:55:03.095404", + "end_time_simulation": 9.979999542236328 + }, + "299fe66074144eeeae74735cf42d161b": { + "parameters_hash": "c8c2849f9ced3d67e82fa39f6b71ffbd", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-02T16:55:07.136974", + "end_time_simulation": 9.979999542236328 + }, + "f35a240813de54f29dc888782a762b5c": { + "parameters_hash": "3289ee2a544b2b474c8df97b6dd291c0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-02T16:55:10.121531", + "end_time_simulation": 9.979999542236328 + }, + "115036fcb4cb40278192aeaec3c948ba": { + "parameters_hash": "a3d845d0ffd80def20e5129bfe8a7ea2", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:08:17.924740", + "end_time_simulation": 9.979999542236328 + }, + "324adbad09bc4d408c447caabe70fc2a": { + "parameters_hash": "446bfe17f003567a2f310714ab14ddad", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:08:25.331427", + "end_time_simulation": 9.979999542236328 + }, + "8e0ac4cb496a4be388e74d0067e210c8": { + "parameters_hash": "c8311779dfdadd2f7aa7ae6141f2db58", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:08:29.671563", + "end_time_simulation": 9.979999542236328 + }, + "83944351735247408e2aee77dbbf6764": { + "parameters_hash": "dd3b3be56e4693f9e0ae24656b3ced69", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:08:33.572835", + "end_time_simulation": 9.979999542236328 + }, + "12e11728995e5433bf6f916f77df797c": { + "parameters_hash": "3b47d4ec0227e2a024da59169881e080", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:08:37.685296", + "end_time_simulation": 9.979999542236328 + }, + "5307b357050d451288c6abb8a709189c": { + "parameters_hash": "8b22761f559132dd778e73c46456f120", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:08:41.915222", + "end_time_simulation": 9.979999542236328 + }, + "2ad573a641685ff49cf2addfff6a0db9": { + "parameters_hash": "37b0dc20f6cd94de8bdf9cf063381b31", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:08:46.232375", + "end_time_simulation": 9.979999542236328 + }, + "390348c49d7d401f89e9841c0e8e8180": { + "parameters_hash": "3fa7f843bc4232439e9c6beacfa7efd9", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:08:51.116679", + "end_time_simulation": 9.979999542236328 + }, + "f2e07a162667420bb31b9c0e2a4978af": { + "parameters_hash": "66c08bf5eb155e9dca31672edfc40357", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:08:55.003721", + "end_time_simulation": 9.979999542236328 + }, + "9bdcce18b1444027adf2c5ed677ddb0a": { + "parameters_hash": "98646ff9e2d4aeb21815ceecd9db5abd", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:08:58.918571", + "end_time_simulation": 9.979999542236328 + }, + "ca1e7e6ab7dc48e08acf47bee4111bb0": { + "parameters_hash": "4042ed275b58effd8ee3fdc5152663b0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:09:01.599497", + "end_time_simulation": 9.979999542236328 + }, + "d23ec53ba66341efa6843bc6f98387fa": { + "parameters_hash": "c4d6c4de148092a5b64dd55fbf31be83", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:09:04.153294", + "end_time_simulation": 9.979999542236328 + }, + "7909eebb4dd951bdbfb2a65d5567c786": { + "parameters_hash": "204473fca59e8d8c08e89e1e47655a97", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:09:06.416783", + "end_time_simulation": 9.979999542236328 + }, + "c20a35e0910a45f1b0591a92fba24107": { + "parameters_hash": "c17885f796b002fc4a38cef242d991a5", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:09:09.461818", + "end_time_simulation": 9.979999542236328 + }, + "b0af5a89bfe34a11837d6a7c33da14ab": { + "parameters_hash": "c6cdd43b00c553c36af23e5e64a34591", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:09:13.010678", + "end_time_simulation": 9.979999542236328 + }, + "8e7a65f97c824a358a67ca2cc8d5859e": { + "parameters_hash": "ba7d2fd420441ca474f06f6979323825", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:09:17.070106", + "end_time_simulation": 9.979999542236328 + }, + "cb535045069d579cba86fe77080233cc": { + "parameters_hash": "f4a7c4744f5d7092a84ca0ecd179baf4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:09:20.676350", + "end_time_simulation": 9.979999542236328 + }, + "1eac3a502bfb4a10aebbadb4716eb24e": { + "parameters_hash": "370115ca47ca206d79ac2720631ab442", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:09:24.314112", + "end_time_simulation": 9.979999542236328 + }, + "04ba28bc53995217a4e20da250d42737": { + "parameters_hash": "02a21701bfbce680f3ccc2f6f7030c29", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:09:28.367454", + "end_time_simulation": 9.979999542236328 + }, + "3dee5963901d530e9f2fde3c84a55982": { + "parameters_hash": "26468440501aa53665093719ddd3f007", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:19:03.054342", + "end_time_simulation": 9.979999542236328 + }, + "c12cd952cbaf5575a408e326728f2a6c": { + "parameters_hash": "1c9a10dc9ee439f10176803c82bfe4e9", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:19:10.437380", + "end_time_simulation": 9.979999542236328 + }, + "7e7f4c9626bd59b2aec6cb9015bcdbe8": { + "parameters_hash": "e44fb827c53aa5886376b0bc56e03de4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:14.151089", + "end_time_simulation": 9.979999542236328 + }, + "3dc5e2e8d7f150fbaf80df25ba9c426a": { + "parameters_hash": "d0a0934b84401abf6d427c5051cdb5a9", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:19:17.281375", + "end_time_simulation": 9.979999542236328 + }, + "f749d3cc55495bde9601b81bd421a806": { + "parameters_hash": "0967f76e1d42f3aaf99759182d2ca933", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:20.384929", + "end_time_simulation": 9.979999542236328 + }, + "5f11dee28e0854aeacc38846ccd621af": { + "parameters_hash": "849ce93e69fc1b5090a5fea099a287a3", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:19:23.204839", + "end_time_simulation": 9.979999542236328 + }, + "32358cb7106859ad9c09625edc6099b8": { + "parameters_hash": "8ca94a818ac6e8d3b7d3284b59ce86bc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:25.914506", + "end_time_simulation": 9.979999542236328 + }, + "f9ee111c4748596db0073bc71a2fdd08": { + "parameters_hash": "e9ff33741d12af05198c3039093f80e1", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:19:28.035628", + "end_time_simulation": 9.979999542236328 + }, + "d15f45e72ff05e9da39213b18022226a": { + "parameters_hash": "135095b2ca7371a3e997d5093e685afe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:30.223257", + "end_time_simulation": 9.979999542236328 + }, + "6b38e6d31f125b6c9d093a5e9dbb8477": { + "parameters_hash": "a26af54b08ea00b2c63d081720f19da6", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:19:33.586884", + "end_time_simulation": 9.979999542236328 + }, + "7e68a06703d8560086da1e845d8df63e": { + "parameters_hash": "7a25af1cf98d8612f16af4d8f34f2a89", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:36.019627", + "end_time_simulation": 9.979999542236328 + }, + "ee05348f9b6554339e88994b03343d56": { + "parameters_hash": "fa9709fac7dc2fcf2215dedaddfd5825", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:19:38.764645", + "end_time_simulation": 9.979999542236328 + }, + "5b3fbb261e59526b83d678468f541958": { + "parameters_hash": "3b5c94b39a4cac35b25ccbc8086d019c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:41.333896", + "end_time_simulation": 9.979999542236328 + }, + "db3d2f8e64db530881922fede7a866e6": { + "parameters_hash": "df939177c2f8d38c900833fe2b9b3a06", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:19:44.368162", + "end_time_simulation": 9.979999542236328 + }, + "bd4e29caee035f44a9f549102dc9ccbb": { + "parameters_hash": "439f298cbba699ebae508dc1c0dc6f33", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:46.725622", + "end_time_simulation": 9.979999542236328 + }, + "f5797a3004d252b2adafeee929b2b171": { + "parameters_hash": "a6a6718bdbbe24ed77ae47bc23e322fd", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:19:49.149239", + "end_time_simulation": 9.979999542236328 + }, + "511406ea1c4a55b381d2beb95d732218": { + "parameters_hash": "cb2c11523d2c18a695822d7872381b6c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:51.753849", + "end_time_simulation": 9.979999542236328 + }, + "c6bc225ea6a252658393a8398bb623a2": { + "parameters_hash": "c1ca5ec6af78d0b588869f0033214301", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:19:54.966933", + "end_time_simulation": 9.979999542236328 + }, + "e3f343e9b0d159068d9ca528689e110c": { + "parameters_hash": "77299e2b392195162cee16897b5bb8f8", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:19:57.889866", + "end_time_simulation": 9.979999542236328 + }, + "fab7e903269f59f6bb028da27bbc57ac": { + "parameters_hash": "937ec51440271f1db061ce6526758a09", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:20:01.466993", + "end_time_simulation": 9.979999542236328 + }, + "1c73d800d9315549b771c5877dcffe63": { + "parameters_hash": "6dcd021ee4bf6795dca12bc0fc62485b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:05.322351", + "end_time_simulation": 9.979999542236328 + }, + "7d1520b2be3c531ab4acbe645c9e18cd": { + "parameters_hash": "4cfa3a4e5d448cf4fa577012c45d555f", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:20:08.741013", + "end_time_simulation": 9.979999542236328 + }, + "62ed04ce3cd053289994abb15678c5a3": { + "parameters_hash": "0fbb57a2080b2dfdd34b44d50aaf69db", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:20:13.296135", + "end_time_simulation": 9.979999542236328 + }, + "8d6c0adc482e5a7bbc3b83b67a50d4fa": { + "parameters_hash": "e03ce5e88e68cb15067587f94c72a8d7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:16.455900", + "end_time_simulation": 9.979999542236328 + }, + "402307421702585c86a2ff40516deca2": { + "parameters_hash": "8a106e0f4c530f51c12d5dd98af10671", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:20:20.300425", + "end_time_simulation": 9.979999542236328 + }, + "52f31b9a29ff51da8a545ea0c4c1f72c": { + "parameters_hash": "6c9e804d750e2dd33fa7fa245cc17c10", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:23.785300", + "end_time_simulation": 9.979999542236328 + }, + "f6d9750d47895106b29d9a44193f66c0": { + "parameters_hash": "77178372b97f0b37efdd1ab61bdffe94", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:20:27.072081", + "end_time_simulation": 9.979999542236328 + }, + "949fa7f1d91f51db988fdbba1f614fa0": { + "parameters_hash": "b3311995fe61c6a6845ff392e153b665", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:29.894504", + "end_time_simulation": 9.979999542236328 + }, + "0d79edc07243565d8d8b5380ef81e340": { + "parameters_hash": "80bc81c5eb48eb5d2ec18c66f060fe46", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:20:32.594121", + "end_time_simulation": 9.979999542236328 + }, + "f1feff149da450fdb736d5eb8bb67861": { + "parameters_hash": "6aeec28d74cf3db6bfd28d45b72afdd6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:34.812921", + "end_time_simulation": 9.979999542236328 + }, + "b167d8b014d350138cd4f321d3739019": { + "parameters_hash": "6e348ce64c806ba062f06165b352aaa7", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:20:37.784008", + "end_time_simulation": 9.979999542236328 + }, + "5c172c6354ab5ababc9ad4a31d6967d0": { + "parameters_hash": "8f9533cd884f8458582cf199d5fc7dfe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:40.427473", + "end_time_simulation": 9.979999542236328 + }, + "6011cc573aec59568fe589231a1d8afc": { + "parameters_hash": "017622fd73ceffac6364f4a21d2c5d62", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:20:43.344279", + "end_time_simulation": 9.979999542236328 + }, + "a1cc58ca322f5e498e06cbf6e464b1b2": { + "parameters_hash": "a2cea67975e5ae394e15f335dc2f1d0a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:45.684355", + "end_time_simulation": 9.979999542236328 + }, + "aae536d9bb9257a495ff15a63cb8d7fc": { + "parameters_hash": "466ef748d6283cca46900a2c517637a4", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:20:48.361551", + "end_time_simulation": 9.979999542236328 + }, + "e04cb4ae5f875bea884b9884478062b7": { + "parameters_hash": "f0b7749d76f396972b8778a95eb5acd0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:50.779098", + "end_time_simulation": 9.979999542236328 + }, + "a4ac4f8f3c1e58f6bdb0e745fa6542a3": { + "parameters_hash": "163aa838809ee42b0fcefa6cbea49b8b", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:20:53.820551", + "end_time_simulation": 9.979999542236328 + }, + "66445369e80a5305bbc0ead043ba7032": { + "parameters_hash": "d92a0b926a2825ae6cff680651138c6a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:20:56.661015", + "end_time_simulation": 9.979999542236328 + }, + "6ad0498cacfa5d2b93dc8200c2a49d78": { + "parameters_hash": "4116cc5a3a2e1d4bda730cc7f253ed63", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:20:59.692355", + "end_time_simulation": 9.979999542236328 + }, + "7123f5d4d81d55deb3f8d8bff6ce987a": { + "parameters_hash": "0d54d1e9f40f89129f03e901ad2ba210", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:02.113920", + "end_time_simulation": 9.979999542236328 + }, + "310624e9965955a6a18b6b834ae98a9b": { + "parameters_hash": "360e6580b30bbead883edd4f141897ae", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:21:04.770374", + "end_time_simulation": 9.979999542236328 + }, + "4205904bfd8656028c2195fc11ccd587": { + "parameters_hash": "df5bde9ed7a74e81b2533332ba8b86e7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:07.136269", + "end_time_simulation": 9.979999542236328 + }, + "23cca9a4b30252e89b56e757b3b6b0f0": { + "parameters_hash": "29523f34cd781b060565e7e04af96b5c", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:21:10.133252", + "end_time_simulation": 9.979999542236328 + }, + "cc910fa7368c5587804f005866a22079": { + "parameters_hash": "0e8e00c54151068a7896ec46ef19000e", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:21:13.750392", + "end_time_simulation": 9.979999542236328 + }, + "3a0ba666136056dcb44006d2e7767b72": { + "parameters_hash": "35c7c2c95ecb9343403338b96fbe0043", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:15.974489", + "end_time_simulation": 9.979999542236328 + }, + "e38d6ac61c9754ee8ff6d3390dd3cb25": { + "parameters_hash": "6b31e0e7e679c910b1ccd6944fa43332", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:21:19.568691", + "end_time_simulation": 9.979999542236328 + }, + "8246b7dd00fe5e31b97edb07cff0b941": { + "parameters_hash": "9b876e66c58dda02fb9b7abeec7f1fbd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:22.781745", + "end_time_simulation": 9.979999542236328 + }, + "39112af1e7ce59d6b1e3d093ca15963c": { + "parameters_hash": "b3ff9872624de772c87006c5321a70b2", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:21:26.379199", + "end_time_simulation": 9.979999542236328 + }, + "c47fad1640ba5de48aed4a0249c52753": { + "parameters_hash": "3f7955d21b09b8090c26bf65ef188eed", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:30.788928", + "end_time_simulation": 9.979999542236328 + }, + "03e7187b9a695b87b9b91e652c2c9ec8": { + "parameters_hash": "2534b1b00fe5722ba30366c9b9ae5f23", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:21:33.864708", + "end_time_simulation": 9.979999542236328 + }, + "e99234eec94d5275a0df8e2ee865d42e": { + "parameters_hash": "96eeb3908fe98ccf2791f9f6087dc67f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:38.031548", + "end_time_simulation": 9.979999542236328 + }, + "a1e990b8d4635985a203b23e9c5d537e": { + "parameters_hash": "e18ae83c535faca16c50e6267099a38f", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:21:42.573366", + "end_time_simulation": 9.979999542236328 + }, + "fc8171e88550540ab2c17990d121cc4e": { + "parameters_hash": "7786dc66bbf849d815d3ecfb7c7fb021", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:45.839868", + "end_time_simulation": 9.979999542236328 + }, + "7c0aa99351d15f6483892c4bad9c34e1": { + "parameters_hash": "d4c1c6b86ae7e90d99976530f10dfdc0", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:21:48.687638", + "end_time_simulation": 9.979999542236328 + }, + "ed4c1f47ee5f5af5bbda2a7d57ae9d1d": { + "parameters_hash": "aaf89c8c165ceb1d270da8b536b08757", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:51.571877", + "end_time_simulation": 9.979999542236328 + }, + "fcb1147346ed5f488d0e336e7dc846ce": { + "parameters_hash": "2310f88d77b0de8eb33310de2b97dcd3", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:21:54.533690", + "end_time_simulation": 9.979999542236328 + }, + "ae4cbc4ad8a750cca5534f2bca0ec872": { + "parameters_hash": "7a6b357616158ff8cd3a39bd9158424a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:21:57.048102", + "end_time_simulation": 9.979999542236328 + }, + "87de692370f95c8389ed84bbe1406168": { + "parameters_hash": "62ed0b58acdb76d109e7b8a7f7f81d2e", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:21:59.726811", + "end_time_simulation": 9.979999542236328 + }, + "2f685fe08a6852f4a6c63b7ea5dfa1e3": { + "parameters_hash": "c0592f11a5bdfd0f63d5a95955d92647", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:02.158082", + "end_time_simulation": 9.979999542236328 + }, + "66299bc3547353e08fba908f849f0db6": { + "parameters_hash": "8af14367f8009822de95853bee390ff3", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:22:05.113037", + "end_time_simulation": 9.979999542236328 + }, + "af94ec34887a5450a2083cd0420e684c": { + "parameters_hash": "03f8bd95f820e740da1de183422297bf", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:07.789981", + "end_time_simulation": 9.979999542236328 + }, + "f343283be5805a04944b0a508b5150ad": { + "parameters_hash": "ea0da5ebd1201bb8e3c9ee70107ba26f", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:22:10.776430", + "end_time_simulation": 9.979999542236328 + }, + "3a2522adb31254eeb45ef0f5c13b4a4e": { + "parameters_hash": "c90b47885b0ef372ca71d9b0fea18571", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:13.584927", + "end_time_simulation": 9.979999542236328 + }, + "1e4d0d6cd30859d491b34dc0200b99a7": { + "parameters_hash": "1ab4e214d62af7d7f87e0dcb0ebf3f3b", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:22:16.061216", + "end_time_simulation": 9.979999542236328 + }, + "e0934b52a37759e3a90cd51c68f250e5": { + "parameters_hash": "5cb01e446967ea0520dfbbf1e92eb801", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:22:18.323702", + "end_time_simulation": 9.979999542236328 + }, + "87325c60c6b652c68a94952e513ead83": { + "parameters_hash": "48302b91e7fe5ed34d50cd23b8a32157", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:20.569917", + "end_time_simulation": 9.979999542236328 + }, + "c5d7959f17da52dda612bcc328e0b427": { + "parameters_hash": "8fe02ddec1e7c8b13f5eabdd5fa43280", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:22:22.876186", + "end_time_simulation": 9.979999542236328 + }, + "5f4ca002e3e552fd843e65da5d32624b": { + "parameters_hash": "44f20dca16e0b58aa760f6b27f78f50e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:25.480563", + "end_time_simulation": 9.979999542236328 + }, + "5d6a4a65d8055cb1a1f66911d6aeee45": { + "parameters_hash": "38ede8634b3bc3f6f71b7a72c6d002d8", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:22:28.578608", + "end_time_simulation": 9.979999542236328 + }, + "4e7c2c03164851f8b818421a1c12e7b8": { + "parameters_hash": "7b43b8c265eafcad9323009a45fcca12", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:30.958268", + "end_time_simulation": 9.979999542236328 + }, + "54558f7c9e4156c89faab7fdc3c9915f": { + "parameters_hash": "313f3233847b556aead7398db4fc4681", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:22:34.154349", + "end_time_simulation": 9.979999542236328 + }, + "ea761ed0b5c35063bdca5362e97d1c43": { + "parameters_hash": "6e13d8d5db7cb8c69d712db2c8129c67", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:36.546566", + "end_time_simulation": 9.979999542236328 + }, + "bbd4478022895ab89a416d45dd4cac2a": { + "parameters_hash": "c053d0fb93d2d0e0eef7cde0ec8c2500", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:22:39.333218", + "end_time_simulation": 9.979999542236328 + }, + "a18db7fd10325fbda26dee363b863021": { + "parameters_hash": "1dc70431311ff904ea26e9f405bba0eb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:42.357932", + "end_time_simulation": 9.979999542236328 + }, + "7128017d5b8b5eae81eba5424f15831e": { + "parameters_hash": "59b8cdc435b84e5e7741e9fd32d8c817", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:22:46.073990", + "end_time_simulation": 9.979999542236328 + }, + "dccfe17d13c4585ab949adf38784cd7c": { + "parameters_hash": "c8ea21610a09b1df9c58cefbc7b47f3f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:48.785151", + "end_time_simulation": 9.979999542236328 + }, + "07fd7c347c7354bbb738d7765a2fea14": { + "parameters_hash": "d3cc78c8149b05f11e9d0db3c823cb40", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:22:52.136512", + "end_time_simulation": 9.979999542236328 + }, + "67a31bd5c6bc5f568f8399c101d75c09": { + "parameters_hash": "56b0287f5d937f4d1db9f311d9ee3ba3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:55.023999", + "end_time_simulation": 9.979999542236328 + }, + "3cdf2c46f20e55c0b081c3745fbf8c40": { + "parameters_hash": "08de11802046048c83d048556c02be24", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:22:57.861170", + "end_time_simulation": 9.979999542236328 + }, + "f7811f809b21592fa7a6c70f94b385f2": { + "parameters_hash": "57f605f60ad044805a3060551d2827a4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:22:59.960948", + "end_time_simulation": 9.979999542236328 + }, + "922bcfee14585e8aae6ca505606ca4a5": { + "parameters_hash": "d3225aa28e255cf962dbec33e407cced", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:23:02.453992", + "end_time_simulation": 9.979999542236328 + }, + "9ef73de18677509788383bbf95a7f5ca": { + "parameters_hash": "d28644e10afb56224d16f887e0886312", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:04.971715", + "end_time_simulation": 9.979999542236328 + }, + "877d81a677de50b4bbc2d44a2c0d471c": { + "parameters_hash": "b79085d506f2823b3937d4ee30497177", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:23:07.572650", + "end_time_simulation": 9.979999542236328 + }, + "5475ede1d1f35da68401fda3cc89cd60": { + "parameters_hash": "3bf0e888e2f3714490cd98c8a5701ef5", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:09.799465", + "end_time_simulation": 9.979999542236328 + }, + "941ff11f95d55f1cb9960e09546e010c": { + "parameters_hash": "a637cc351ef4a5579926f1f8823d9b8d", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:23:12.103938", + "end_time_simulation": 9.979999542236328 + }, + "f55f97e4bc675113bcc662046aeced55": { + "parameters_hash": "1d7bea093a0b637d6f581901e16c4873", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:23:14.360885", + "end_time_simulation": 9.979999542236328 + }, + "2cbee75b98bd5d198e48adec329b820b": { + "parameters_hash": "5d61b5a5ce8f77e5d6cd809bbd3168c6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:17.077843", + "end_time_simulation": 9.979999542236328 + }, + "1a078343ddd8560584af232b1e580924": { + "parameters_hash": "4ffcf69227002c12d2b936f4b3766a2e", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:23:19.615889", + "end_time_simulation": 9.979999542236328 + }, + "7aacc34439d1535aae8db548c804f2f6": { + "parameters_hash": "b7657563554d611cd8b5c9d7bad98a0e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:21.895654", + "end_time_simulation": 9.979999542236328 + }, + "62e1ac7a217b53d199fbd04081b5d4ce": { + "parameters_hash": "968c07173a256bb7dcaba5eefa8ce79f", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:23:24.419856", + "end_time_simulation": 9.979999542236328 + }, + "77c64637989c53aea778418fdf6378ac": { + "parameters_hash": "68a875126a22c0bd199a0ed8ee28ca49", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:26.779296", + "end_time_simulation": 9.979999542236328 + }, + "e03b46ffaf135d9385aede10d14ea867": { + "parameters_hash": "5257a4f7a906cc886b9191c447839d97", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:23:29.011034", + "end_time_simulation": 9.979999542236328 + }, + "4487f405ba015e83bbdaf61a46a00eca": { + "parameters_hash": "a8d6ea81aad8083cabba2cd6620ee6be", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:31.067371", + "end_time_simulation": 9.979999542236328 + }, + "1a584ed66ff75019abd1ca91511466e6": { + "parameters_hash": "90f5af6bb8ab6302f06c85097f7f3c68", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:23:33.518220", + "end_time_simulation": 9.979999542236328 + }, + "4788dd4d7ac85020a4c017935d53e9a7": { + "parameters_hash": "5c4a7d6cd3cb80b81a8908e8e7cc429d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:36.140492", + "end_time_simulation": 9.979999542236328 + }, + "1ea1a6d0f2e052cca3d3435c026f1593": { + "parameters_hash": "c44822299c838155efc2da880636c1a1", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:23:38.878885", + "end_time_simulation": 9.979999542236328 + }, + "9327336d7a4e582da7291748cdd951b3": { + "parameters_hash": "e8be4242b4791a8012554ff95ca20744", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:41.362391", + "end_time_simulation": 9.979999542236328 + }, + "04c24e1e18f4559d91b7f4e58b1f301f": { + "parameters_hash": "c65506556e8f22f72d8f69ddb69d42e4", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:23:44.548281", + "end_time_simulation": 9.979999542236328 + }, + "d734036c1d165e11bc334542a03ba75e": { + "parameters_hash": "3e161867661e9d9dd880a996d86d69c9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:23:47.236018", + "end_time_simulation": 9.979999542236328 + }, + "833af482c5f255bc9ba13190d19a98e6": { + "parameters_hash": "4ebf5a8394c86351b82d032be7f79f37", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:23:50.608247", + "end_time_simulation": 9.979999542236328 + }, + "5b7072c9407d53d584947b3af846e1fc": { + "parameters_hash": "8452aefc449794b2b165871050e5fdee", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:23:17.234522", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 1 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: When trying to advance one time step, nonlinear solver failed to converge, residual norm too large.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Here is the set of components with unconverged equations:\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.elements.mass' (line 40)\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.elements.friction' (line 93)\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.translational' (line 19)\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 1 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: When trying to advance one time step, nonlinear solver failed to converge, residual norm too large.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Here is the set of components with unconverged equations:\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.elements.mass' (line 40)\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.elements.friction' (line 93)\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.translational' (line 19)\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 1 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: When trying to advance one time step, nonlinear solver failed to converge, residual norm too large.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Here is the set of components with unconverged equations:\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.elements.mass' (line 40)\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.elements.friction' (line 93)\n \n 'simulink_model/Mass With Friction (PB)'\n Equation location is:\n 'foundation.translational.translational' (line 19)\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "3fbb72b0bff757a0b69db582c8ea79e0": { + "parameters_hash": "88df135b8bde92f03cfc22529712be60", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:24:06.642762", + "end_time_simulation": 9.979999542236328 + }, + "ec1e66da291b5cab987b3cc932354041": { + "parameters_hash": "16ab4f266e92e945aa231b3b8bc11cfe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:24:09.228760", + "end_time_simulation": 9.979999542236328 + }, + "7e497d9804d95bbabb6f80aa0e0d9cd3": { + "parameters_hash": "4e8cc10a61ba1d2ebea54a47b95e4410", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:24:11.386146", + "end_time_simulation": 9.979999542236328 + }, + "bab688668fc25231812410e9d57b9fb2": { + "parameters_hash": "f88575e55e54fc00cdc83757da07ab0a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:24:13.324486", + "end_time_simulation": 9.979999542236328 + }, + "b794f7c284485e548f68445b0c123a3a": { + "parameters_hash": "54bc13209359a56d5ca1757ba140ef8f", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:29:04.307443", + "end_time_simulation": 9.979999542236328 + }, + "3a49223ec7375b1f8dff9acc6d1c19b2": { + "parameters_hash": "b65717870628b30d02646cf3d7b996d4", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:29:09.455264", + "end_time_simulation": 9.979999542236328 + }, + "9f9658ea035d54249f81496713d6175b": { + "parameters_hash": "562c4bf99644b471d91f722f16cb9272", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:12.910226", + "end_time_simulation": 9.979999542236328 + }, + "7c987e2a553e5517805516ef6b6e8949": { + "parameters_hash": "6e87cd8b23338b79539761462a2cd573", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:29:16.660172", + "end_time_simulation": 9.979999542236328 + }, + "a9e4a261435f5a70b3ec58bba1aa1905": { + "parameters_hash": "8d1870e0b3901a7049047162f93c0dc1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:19.002580", + "end_time_simulation": 9.979999542236328 + }, + "1dbde2fdb75b5ca294baa80857327b67": { + "parameters_hash": "8c7ac627e25a9b3e9980414b56764da4", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:29:21.880854", + "end_time_simulation": 9.979999542236328 + }, + "7e83ac69681957faadab02f31d87857f": { + "parameters_hash": "465ef95200726bc481bb50d5cdfdf5bd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:24.389256", + "end_time_simulation": 9.979999542236328 + }, + "47e6a627cba254d88c71253aa4ba4597": { + "parameters_hash": "8c09002d4acff46dc8cb6e54c251a7e8", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:29:27.552015", + "end_time_simulation": 9.979999542236328 + }, + "afbb98f1ebe455ffb3356453e1ab2a98": { + "parameters_hash": "64d5bc39dd4d5018690ef680ffbe388a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:29.890645", + "end_time_simulation": 9.979999542236328 + }, + "85e82eff3cdb5ae48285f4ad955f3bec": { + "parameters_hash": "dc2ab9185e65a7570ea84578fd1714c3", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:29:32.815439", + "end_time_simulation": 9.979999542236328 + }, + "1943479ece285517bda6c1db6220109e": { + "parameters_hash": "e75a8a62cad3d630122b64f3c7b9d657", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:35.578166", + "end_time_simulation": 9.979999542236328 + }, + "c38c99a93c7e5be98eddfba0f44e9e67": { + "parameters_hash": "84dcbddcc995f13c80fbdd51d6d642e4", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:29:38.676019", + "end_time_simulation": 9.979999542236328 + }, + "0328b1a464e85b68ae1bcb7e396a02cc": { + "parameters_hash": "89b8b34b820fab80225d93ac6551defe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:41.062352", + "end_time_simulation": 9.979999542236328 + }, + "83d0974ba2ac557faa7fede0733f17e6": { + "parameters_hash": "71cce5fa461d1b9a7fc34e34699f6aab", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:29:44.076640", + "end_time_simulation": 9.979999542236328 + }, + "b40a37a9cfa05f04b49a46de3c7b4fdc": { + "parameters_hash": "2660a72256ba1e6a72049384318a3809", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:46.656215", + "end_time_simulation": 9.979999542236328 + }, + "6735288b3b0f5325996a923190a9730f": { + "parameters_hash": "c795500a5ba73b6b32f8ccbef204a5d1", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:29:49.559404", + "end_time_simulation": 9.979999542236328 + }, + "770f3e2000b05968ac88420e4bfe0e06": { + "parameters_hash": "c0196fa5ab491beefbd8b10e490a3bde", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:52.193002", + "end_time_simulation": 9.979999542236328 + }, + "4e0bf8aa58a258a29f68903d43263e66": { + "parameters_hash": "b5c684ac546da6ddf5c175e0aab7c36f", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:29:55.077833", + "end_time_simulation": 9.979999542236328 + }, + "4cd18a38903257278079f369d2ad0389": { + "parameters_hash": "10d7ce9996abea8df0f9fff35754cd55", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:29:57.267408", + "end_time_simulation": 9.979999542236328 + }, + "906b2f75d36d5647a8fe4ef05f19fa5d": { + "parameters_hash": "1e8c2a2e489efb1d77918c94cd44b248", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:30:00.530356", + "end_time_simulation": 9.979999542236328 + }, + "fe90c96f6de35202b5694fa9bd27f7b7": { + "parameters_hash": "7a0384f8c3a97aef427ed71659113862", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:03.139387", + "end_time_simulation": 9.979999542236328 + }, + "b4591a30e59f5c4897a79165f3d336ac": { + "parameters_hash": "4b1ec66798afcd1851af4e00efd3a0a0", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:30:05.388752", + "end_time_simulation": 9.979999542236328 + }, + "b1759365df145f45a7021a3250a6e726": { + "parameters_hash": "f1c5215ff3d6b4eb5153374bf007317f", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:30:08.512516", + "end_time_simulation": 9.979999542236328 + }, + "11e31d9b02a55d55b8257c2086996f02": { + "parameters_hash": "096446e6e61046cff462e755c0695df9", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:11.720812", + "end_time_simulation": 9.979999542236328 + }, + "4936a5b652995d1cbb5987f3c7ee6866": { + "parameters_hash": "8bbd58b7e57baf7a421b38324475239b", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:30:14.937041", + "end_time_simulation": 9.979999542236328 + }, + "ef848d0f59d0540bb114a1f34b9226c6": { + "parameters_hash": "269bdfa10fddcf5985724c40fa3e0619", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:17.533972", + "end_time_simulation": 9.979999542236328 + }, + "8b39991d40a95931b072f880580f82e2": { + "parameters_hash": "d98d182980317d49a7880fe55d3b6a27", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:30:20.766223", + "end_time_simulation": 9.979999542236328 + }, + "3c985b4028495470acbf1b2b5096aa43": { + "parameters_hash": "44db16c24456454141f292e6320b89ec", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:23.267634", + "end_time_simulation": 9.979999542236328 + }, + "0c7f3359200b539882010d801f5a4e7e": { + "parameters_hash": "20de7af9ea707d59c237172494362715", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:30:26.025930", + "end_time_simulation": 9.979999542236328 + }, + "885d7a2b19ce5260845974ff79c54cb1": { + "parameters_hash": "cd1785ce32b9b12d4b60b6fe9e35c0dc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:28.284323", + "end_time_simulation": 9.979999542236328 + }, + "bec48ee90a885bfdb8a06c92f88fe341": { + "parameters_hash": "68518416ef161d1ccd2950310a1e488c", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:30:31.245117", + "end_time_simulation": 9.979999542236328 + }, + "7ae25c8485fc5573b4763109788b7afc": { + "parameters_hash": "b2eae6f8dc7a55a96ec2576e4f375446", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:33.319342", + "end_time_simulation": 9.979999542236328 + }, + "d862f707f8df5007ae7dceda27542f8e": { + "parameters_hash": "e35d328447edbe2f453e27e0e51576f8", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:30:35.904685", + "end_time_simulation": 9.979999542236328 + }, + "fd958f503fd95672a59e0af21eb573c1": { + "parameters_hash": "22741275e36148a6ffa005a4f326fb76", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:38.515778", + "end_time_simulation": 9.979999542236328 + }, + "6d536e8bdbac58d5bf6f6b49e9d4809c": { + "parameters_hash": "97f839d4adcf98b4459d1718e3a906c7", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:30:41.101087", + "end_time_simulation": 9.979999542236328 + }, + "1200e3b0d11b57208efe4757865d907c": { + "parameters_hash": "a475275a46ef7cec8455d42ec7006554", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:43.822989", + "end_time_simulation": 9.979999542236328 + }, + "38a0490d29e35d33a2413a8e91a01a99": { + "parameters_hash": "483d036fc882acd5e6f9744180b9136a", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:30:47.002292", + "end_time_simulation": 9.979999542236328 + }, + "c4de817969c250eeb0ea45d9b05a7bd4": { + "parameters_hash": "e51ef7c8087f9c54621a7de4cd6beb63", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:50.074761", + "end_time_simulation": 9.979999542236328 + }, + "d35f3d04b1d45732a0995b7c11b6e319": { + "parameters_hash": "a1d1fd2a5b9ab80d3c0aed81f1009640", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:30:53.224963", + "end_time_simulation": 9.979999542236328 + }, + "2231a33dd7945a209469b517d09c92fa": { + "parameters_hash": "7eda5226bf751f2383d55ef726136cbc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:30:56.472251", + "end_time_simulation": 9.979999542236328 + }, + "9b1982d86e3b5bb892c7d1270f5e8e4c": { + "parameters_hash": "a2b17a57de69fac673dac3179ab5c83d", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:30:59.449915", + "end_time_simulation": 9.979999542236328 + }, + "b31dd221f61550f0921795527f8937c5": { + "parameters_hash": "2d6372125ade17c06c255aec932ecca7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:02.335846", + "end_time_simulation": 9.979999542236328 + }, + "d7efe6d51f6c5e6d8edd9ddff98f01f4": { + "parameters_hash": "74bef3d968bf0b16487ed0f11716adec", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:31:04.791498", + "end_time_simulation": 9.979999542236328 + }, + "ffcb3e01741a516091f8da7aeb689274": { + "parameters_hash": "ec12dafabe177c85eb913aeb72956169", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:31:07.567330", + "end_time_simulation": 9.979999542236328 + }, + "0c73a0da75c457bf92c8be8ead00be77": { + "parameters_hash": "802d849ede54cc51dbda749a1d5953b0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:09.891628", + "end_time_simulation": 9.979999542236328 + }, + "837041204d635514894a4001a5399e08": { + "parameters_hash": "b7cef4f233063595731a3b3fdbc0e1d1", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:31:12.402647", + "end_time_simulation": 9.979999542236328 + }, + "82a7473e8421563e90b99883b60e3c85": { + "parameters_hash": "b953c99a37bd9c28502c82856d93af7c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:15.158075", + "end_time_simulation": 9.979999542236328 + }, + "f5aa7878f7cb5c18a915763f365d91b2": { + "parameters_hash": "fd1f654de9eb3ed960e2fb35227ae0c5", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:31:17.963889", + "end_time_simulation": 9.979999542236328 + }, + "7f3a8a0a0c53598eac4d7758f59bd3e7": { + "parameters_hash": "5088c7bcaf08ed3bb04cc9d0ea6f2da1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:20.526187", + "end_time_simulation": 9.979999542236328 + }, + "e490a47b88ca5878ac4f0394118d6b7d": { + "parameters_hash": "41df2abd18df561d7f012ad6c414cb21", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:31:23.675355", + "end_time_simulation": 9.979999542236328 + }, + "3b18deeeb83458359024322a970ede3c": { + "parameters_hash": "4f128c1e178043c0d4c41a942bbd7059", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:26.728915", + "end_time_simulation": 9.979999542236328 + }, + "38a3d07744b0580baa0a1f1238c8beee": { + "parameters_hash": "590d6d402baa81a62876d6e1e8f8f0da", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:31:30.197034", + "end_time_simulation": 9.979999542236328 + }, + "b8cc5378594c554eaa42ea0d28cb9956": { + "parameters_hash": "bfaa9cb917a120d3f3ca236454dacdb6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:32.760119", + "end_time_simulation": 9.979999542236328 + }, + "dc27a6bbcb815a968a01e4fecfc5792a": { + "parameters_hash": "401686c902fc35b234deb0f96a6fe644", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:31:35.850148", + "end_time_simulation": 9.979999542236328 + }, + "2df63544473d5f84a1c41fb237a6a313": { + "parameters_hash": "2a4a67500e13d2ff11c3f0315e565734", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:38.583736", + "end_time_simulation": 9.979999542236328 + }, + "9c2d4063dd2e5e38909c6cfa579988ca": { + "parameters_hash": "a7de41d36a6d1b413a0a4a9062a97f46", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:31:41.843306", + "end_time_simulation": 9.979999542236328 + }, + "d9a0f919c0a650b39005599f5cf05361": { + "parameters_hash": "872c4207f0cd741894e2d0cb6ca47064", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:45.020928", + "end_time_simulation": 9.979999542236328 + }, + "a8204a70455f51dc8ef3f57d07a2931c": { + "parameters_hash": "d8347c172caee371da447a11104a27de", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:31:47.643670", + "end_time_simulation": 9.979999542236328 + }, + "a97b441c5682580f85d7e5ede85a9d3d": { + "parameters_hash": "43656794d1d8fca8152e557f1069fe51", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:49.707625", + "end_time_simulation": 9.979999542236328 + }, + "295ebf7ea0855a90bf54a27dc953fdcb": { + "parameters_hash": "e74673ceb326c87f5f762585e5b55f8c", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:31:52.404051", + "end_time_simulation": 9.979999542236328 + }, + "e9389e5b010a5b4dbf0488bfcea65524": { + "parameters_hash": "2e86d5b4e0469ed8aa250db04631d857", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:31:55.114095", + "end_time_simulation": 9.979999542236328 + }, + "52d91d5389a652baa478069ea293ca73": { + "parameters_hash": "695ec75f3338a7ed84dbeffd50c2298d", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:31:58.735215", + "end_time_simulation": 9.979999542236328 + }, + "e64b5fd264805d429951ee510b37273f": { + "parameters_hash": "000af2ac7d7b0519d49ccfed9a64a0b4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:01.030253", + "end_time_simulation": 9.979999542236328 + }, + "0dc61f09ac3a57cba3e986abd378a58c": { + "parameters_hash": "101e41519137ae547190419f525c4d94", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:32:03.606816", + "end_time_simulation": 9.979999542236328 + }, + "799ce85835875f42a8764a56c9360b4c": { + "parameters_hash": "c623b486f6ce3546c00a5a287ab996f0", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:32:06.441845", + "end_time_simulation": 9.979999542236328 + }, + "65c2b04f903453c18e64bb92ef130a14": { + "parameters_hash": "a69907dd6df4f871cd058c304ed94bfe", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:08.846440", + "end_time_simulation": 9.979999542236328 + }, + "8f01ddabf2a25e0fb895bcf5740b74c7": { + "parameters_hash": "eea751c91f4c6140966cc196a8a7bbff", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:32:11.555386", + "end_time_simulation": 9.979999542236328 + }, + "8886f247f5fc51269070d591fedda3b0": { + "parameters_hash": "15a57ff672e33078981550586e0e1828", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:13.706519", + "end_time_simulation": 9.979999542236328 + }, + "d950be4ceaf855e4b99175c9b68e37f5": { + "parameters_hash": "1d0e6402134fc1093ad58701a2c95083", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:32:16.325153", + "end_time_simulation": 9.979999542236328 + }, + "a8a4971544c55e118b8952d77c955c14": { + "parameters_hash": "d67cbf9a4eb31952cff136fceee54185", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:18.880537", + "end_time_simulation": 9.979999542236328 + }, + "0bdced522c2f52eda6ae831b7fbee12e": { + "parameters_hash": "c9c396ba73c83a0ee01b3dce88a6e762", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:32:22.265030", + "end_time_simulation": 9.979999542236328 + }, + "6549b27178fb51a7b3aa837b6420d93d": { + "parameters_hash": "c6cf5aaedd7d764e4a82035f087b48b4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:25.135825", + "end_time_simulation": 9.979999542236328 + }, + "38c4522862295dbab79dfb5f026e39e1": { + "parameters_hash": "682505785e348470f2bd7ea49a8c9640", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:32:28.014303", + "end_time_simulation": 9.979999542236328 + }, + "bebe523c040353a99432652988e6784f": { + "parameters_hash": "fac4a485c86c142023d5028355ce7c9c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:30.785046", + "end_time_simulation": 9.979999542236328 + }, + "e09d2c5fc0795973a13d629404d55356": { + "parameters_hash": "7ab1b2a7e60f356ddda212337fff3295", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:32:33.351569", + "end_time_simulation": 9.979999542236328 + }, + "2bb44e8d4d9e5e0d917a08478fb5cd96": { + "parameters_hash": "182ace317031826035d4ee0a9fe06857", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:36.428633", + "end_time_simulation": 9.979999542236328 + }, + "0655d91288195e93a7903500183734dd": { + "parameters_hash": "3aad474e508a4332f1cbea5e592f08ea", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:32:39.343702", + "end_time_simulation": 9.979999542236328 + }, + "55b112455777579683354ea87b79963c": { + "parameters_hash": "e750ca3a8d0457e9debca34c6f4616dd", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:41.472577", + "end_time_simulation": 9.979999542236328 + }, + "f2f13ebd94865bf28ae7eeea4ef29cdf": { + "parameters_hash": "4a58e5b8d56eafc0f928c6e16c643003", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:32:44.500614", + "end_time_simulation": 9.979999542236328 + }, + "8ed8197c1cb4504db077add780afecee": { + "parameters_hash": "a146e27303f18ec1568b7a0deddc86d0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:47.425843", + "end_time_simulation": 9.979999542236328 + }, + "9d751e87415c503d9e7a9f16d7ee3352": { + "parameters_hash": "de86b297334e2ff6c2690a2df5e09e83", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:32:50.857124", + "end_time_simulation": 9.979999542236328 + }, + "169063bb217959dd93965d3cbf0b3b1e": { + "parameters_hash": "498e4e0ce4fe066dadaffb3787170edc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:53.711904", + "end_time_simulation": 9.979999542236328 + }, + "69697cb6337b531bbaf9fac63a8d002e": { + "parameters_hash": "0d9aa7d26f07197b0911ba1505eebde0", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:32:56.545347", + "end_time_simulation": 9.979999542236328 + }, + "74a2e59ac2af5ac78bdb49211a253236": { + "parameters_hash": "e5afc2239717924401c2ddfdc763d892", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:32:58.883986", + "end_time_simulation": 9.979999542236328 + }, + "2dec7910f3a25f448d279e1bcd61c763": { + "parameters_hash": "9f7d4de458d1c5a1a9e72004056e2663", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:33:01.314991", + "end_time_simulation": 9.979999542236328 + }, + "2e0613b9a1095838aff88b9ee59705c6": { + "parameters_hash": "7c810f66bbc5c03742d1b1b3d262609b", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:33:04.229383", + "end_time_simulation": 9.979999542236328 + }, + "3e136fe7dd465973860febc53d2831db": { + "parameters_hash": "0e4c4d3d94bcd785696111b1d586f8fa", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:06.692335", + "end_time_simulation": 9.979999542236328 + }, + "a5f3dcfdbfd65d64a37157f61e725aeb": { + "parameters_hash": "eb44b1184704c5297677103b659cafc2", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:33:10.214385", + "end_time_simulation": 9.979999542236328 + }, + "909f24cfbbdd58b191c2b1fa9171614f": { + "parameters_hash": "cebe381b049d6c4ac82eaefb6c881a91", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:12.754090", + "end_time_simulation": 9.979999542236328 + }, + "2fc73742c4235516a326bc749d661972": { + "parameters_hash": "717393ab4afb9d948c8faa1abeb944d5", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:33:15.682541", + "end_time_simulation": 9.979999542236328 + }, + "3fd91023bfa2587e934d6c059db0b40b": { + "parameters_hash": "0a24d00c4e6a7f00bc993976c893478e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:18.407961", + "end_time_simulation": 9.979999542236328 + }, + "42ca4e2bcaae5ec098e5ad4b222ac281": { + "parameters_hash": "1f420d33c49c3acd0d17dbd1938ad0f8", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:33:21.113913", + "end_time_simulation": 9.979999542236328 + }, + "8a15799a986450e09b5128bba9bd5926": { + "parameters_hash": "b1d55ca912899361608bbf4e7fe82675", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:23.830436", + "end_time_simulation": 9.979999542236328 + }, + "2b86b4d94caf591e87174426cc139173": { + "parameters_hash": "8b05afe650d8d1f502607383e4e76c5d", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:33:26.957868", + "end_time_simulation": 9.979999542236328 + }, + "67346c6af5ed59e796193080fbb7f671": { + "parameters_hash": "e79228eaf28db8f971152ae9adfffa2f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:29.852959", + "end_time_simulation": 9.979999542236328 + }, + "205b4aca3c8553708418d9c99869d68c": { + "parameters_hash": "6f77b6cd375f286c92bd97ae9d47e5ef", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:33:32.603273", + "end_time_simulation": 9.979999542236328 + }, + "4efb306b01bb5e399ef7ec53ba4b93ca": { + "parameters_hash": "14b4ee7246c68bd24df7fa18d1bef335", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:35.399287", + "end_time_simulation": 9.979999542236328 + }, + "90e7d8773a2b5374bdbb00f284482882": { + "parameters_hash": "44023977f90a8f4b7250ac56e8c021f7", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:33:38.752893", + "end_time_simulation": 9.979999542236328 + }, + "150eb2b34131568c8f01a08b0f409030": { + "parameters_hash": "a25d8d04021e9933dde473b205959241", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:41.547770", + "end_time_simulation": 9.979999542236328 + }, + "b7d2dbe41aef52c58ccf2a7fd13903a5": { + "parameters_hash": "75c5529a24c62c82a061c4a13b78faaa", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:33:44.727209", + "end_time_simulation": 9.979999542236328 + }, + "6753446291a851e293d112309bc7a61c": { + "parameters_hash": "097e2f71301f6d3d70faeec5761ee7c7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:47.226648", + "end_time_simulation": 9.979999542236328 + }, + "4ee166cf23f35c59975b47b8f8e7f243": { + "parameters_hash": "e2490614b001a37aaa5145dfe4524819", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:33:50.679705", + "end_time_simulation": 9.979999542236328 + }, + "f8aa828046535aaa9696acb16b8a4472": { + "parameters_hash": "f52b8f4940ba26cfec4ce7200466030a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:53.201854", + "end_time_simulation": 9.979999542236328 + }, + "9d21dfa3cf015fc18a145501bc1d924f": { + "parameters_hash": "3e2f6592b899fc1e9cf01bd367dc21ea", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:33:56.334160", + "end_time_simulation": 9.979999542236328 + }, + "e08b8ec0e6325d888a6a7e9cdb27783a": { + "parameters_hash": "3d6948aaaef6000818c66109f4b9fc11", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:33:58.852844", + "end_time_simulation": 9.979999542236328 + }, + "e803fdb9e638552d9050b896808b9b46": { + "parameters_hash": "258ce11aaf148879738d75ac5c66e08a", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:34:01.439867", + "end_time_simulation": 9.979999542236328 + }, + "c9fc286b32ce5c6ea4de96c5e2e9e609": { + "parameters_hash": "dcc986db6910d139e37d4b4d4a329c36", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:34:04.719221", + "end_time_simulation": 9.979999542236328 + }, + "e02e03c135665702be545c9c762d63c2": { + "parameters_hash": "34946698e1da896e3a24e9d8aaa9e7dc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:34:08.144930", + "end_time_simulation": 9.979999542236328 + }, + "c6cc5bbcb54c5d37b6f62836386f8f14": { + "parameters_hash": "2d260d089efc673bb4978b5b9318fc7f", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:34:10.861647", + "end_time_simulation": 9.979999542236328 + }, + "a660460a01645f0abac2f12cdb6bc71f": { + "parameters_hash": "3b3d43036aa5cdb3bdd0e6d234689642", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:34:13.119290", + "end_time_simulation": 9.979999542236328 + }, + "240b88ab2e675a53b19085bb1bbb4954": { + "parameters_hash": "d2ccb047ffa972ffd343928375bf9d03", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:34:15.550191", + "end_time_simulation": 9.979999542236328 + }, + "7794b77802cc5321ad547b475dff3d0e": { + "parameters_hash": "800723a8ceb2b370af7b331b28c8bce6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:34:17.848172", + "end_time_simulation": 9.979999542236328 + }, + "5b0516f2dad054e8a7f92d17d94f073a": { + "parameters_hash": "6586c4b5fabee07c37f441c2d9ddf2f0", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:34:21.066642", + "end_time_simulation": 9.979999542236328 + }, + "d434e663c7295a689243bb7538667a0e": { + "parameters_hash": "01534e7809e9210120cb947769f1833d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:34:23.757764", + "end_time_simulation": 9.979999542236328 + }, + "c915d923d7005dfaaf3a92ddd3e68d75": { + "parameters_hash": "0e32bdd238c133adcb18ed1a889477d0", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:34:26.917569", + "end_time_simulation": 9.979999542236328 + }, + "df16ef941ac552b6befe229109cfa7ac": { + "parameters_hash": "5b2819a384739b9231cefb2259f69883", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:34:29.965434", + "end_time_simulation": 9.979999542236328 + }, + "0db3f68fee99504eb8641776494b044d": { + "parameters_hash": "b6204dd5a2cc731460c999588cbc68e5", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:34:32.979898", + "end_time_simulation": 9.979999542236328 + }, + "af0d841e3c595d249fe38c90dfae9434": { + "parameters_hash": "8ee249115f114acd2a5d9b8213312116", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:34:35.622918", + "end_time_simulation": 9.979999542236328 + }, + "7dab5d3431405b48ba484ebb00e67a1c": { + "parameters_hash": "36b52cae240579914cdda0ff0024d46d", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "failed", + "timestamp": "2026-05-05T19:23:29.252988", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 186, in sim_the_model\nSegment: 2 failed. Error using sim_the_model (line 183)\n['simulink_model/Solver Configuration']: At parameter initialization, one or more assertions are triggered. See causes for specific information.\nCaused by:\n Error using Simulink.Simulation.internal.DesktopSimHelper\n Coulomb friction coefficient must be less than or equal to Breakaway friction coefficient. The assertion comes from:\n Block path: simulink_model/Mass With Friction (PB)\n Assert location:\n o In between line: 62, column: 5 and line: 62, column: 11 in file: foundation.translational.elements.friction\n o In between line: 76, column: 15 and line: 83, column: 42 in file: foundation.translational.elements.mass_with_friction\n \n \n Error in Simulink.Simulation.internal.DesktopSimHelper.sim\n \n Error in Simulink.SimulationInput/sim\n \n Error in sim_the_model (line 183)\n tmp = evalc('so = sim(si);'); %#ok\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n" + }, + "a3d4298160c95986b47107a7994ddc55": { + "parameters_hash": "bb611ccdbffa21c079da16d92aff9d9e", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "failed", + "timestamp": "2026-05-05T19:23:30.720123", + "error": "File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 912, in run_simulation\n result = future.result(timeout=_SIMULATION_TIMEOUT_SECONDS)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/futureresult.py\", line 62, in result\n return self.__future.result(timeout)\n File \"/csem/divr/users/tbe/repo/tsENV/env/lib/python3.10/site-packages/matlab/engine/fevalfuture.py\", line 76, in result\n self._result = pythonengine.getFEvalResult(self._future,self._nargout, None, out=self._out, err=self._err)\nmatlab.engine.MatlabExecutionError: \n File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate/run_pending_sims.py\", line 1189, in _run_model_dir\n simulation_result = simulation_api.simulate_recipe(\n File \"/csem/divr/users/tbe/repo/tsENV/shared/simulation.py\", line 849, in simulate_recipe\n all_signal_dict = sim_module._simulate_case_to_signal_dict(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 1071, in _simulate_case_to_signal_dict\n res = run_simulation(\n File \"/csem/divr/users/tbe/repo/tsENV/workflows/simulate_core.py\", line 972, in run_simulation\n raise MatlabSegmentFailure(\nshared.matlab_runtime.MatlabSegmentFailure: File /csem/divr/users/tbe/repo/tsENV/models/simulink/InclinedPlane/sim_the_model.m, line 129, in sim_the_model\nError compiling Simscape network for model simulink_model.\n" + }, + "98a6f68601a95766bdb5dd10773858d7": { + "parameters_hash": "24e5b5544d1c292fde2d234f5bd7c368", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:34:52.517858", + "end_time_simulation": 9.979999542236328 + }, + "44c54da7d4ec52eaaf3f8b2acee5a06e": { + "parameters_hash": "14dbb7a0db83c0530a72cc6cbd0e130f", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:34:55.249246", + "end_time_simulation": 9.979999542236328 + }, + "f0111ccf22595613910b73335bb54a57": { + "parameters_hash": "07f3926bffa1e5c19923a2ff5b920d2a", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:34:58.166069", + "end_time_simulation": 9.979999542236328 + }, + "19d5558897295c8a8e5c36aa81defcfa": { + "parameters_hash": "d12d27703636c120953964c8b2efa672", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:00.785803", + "end_time_simulation": 9.979999542236328 + }, + "54e65627643656659a199ae08c0911bd": { + "parameters_hash": "ed72fa78de23a195dc6e382786a9ec9a", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:35:03.505131", + "end_time_simulation": 9.979999542236328 + }, + "f0abe4876c4d5c9891afdff138b8dec4": { + "parameters_hash": "0f7f1a940a08b57a8caae6ca23769390", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:06.310677", + "end_time_simulation": 9.979999542236328 + }, + "660fdef92d435e0a88bb41b9e958b990": { + "parameters_hash": "bdad9f29b10b1d4f71121b26b315c1ab", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:35:08.995311", + "end_time_simulation": 9.979999542236328 + }, + "d035602158135a218d2b66e09db01f0d": { + "parameters_hash": "5de63953ba0ee0884e65cec0d25f6ffa", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:35:12.021211", + "end_time_simulation": 9.979999542236328 + }, + "1d7d96a1d8dc54e68f7f86e0d013e9c4": { + "parameters_hash": "b84fb521a5a4ee386837d943035caa3c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:14.815546", + "end_time_simulation": 9.979999542236328 + }, + "d432829e0ca05e7880e10413cbb87329": { + "parameters_hash": "3aa4a149407699dae07383fbea0ceb64", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:35:17.727297", + "end_time_simulation": 9.979999542236328 + }, + "bc11486fcbbb57ff9e3cc83c7b79bf46": { + "parameters_hash": "1f9e4a2081eb684f5d1fad791cdaded6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:20.513715", + "end_time_simulation": 9.979999542236328 + }, + "8aa929de582d5cbf83aa53fe9ab0a174": { + "parameters_hash": "da17d9e56c653e56ec9675171884e429", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:35:22.761687", + "end_time_simulation": 9.979999542236328 + }, + "7da48ac0612d5a6abc2e38002c666859": { + "parameters_hash": "96b30b4d06b466f9b3f2e2efa5855ae6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:24.806191", + "end_time_simulation": 9.979999542236328 + }, + "fd18b1a869a557619f11c8416a1cb748": { + "parameters_hash": "69970ec85453e3f3155876cecb0a58c1", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:35:27.187686", + "end_time_simulation": 9.979999542236328 + }, + "82e85ddca98050c2b5ed32d8f7bf4f82": { + "parameters_hash": "8ec1e1a50dbe77111d05c831bf97671d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:29.879549", + "end_time_simulation": 9.979999542236328 + }, + "e7dc350bb75b59679e84a7a4b4fe6a5d": { + "parameters_hash": "9bb3d22b07b45c42aec880c01538968e", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:35:32.777121", + "end_time_simulation": 9.979999542236328 + }, + "6a66ed579c555772ae05a24bfd60ec39": { + "parameters_hash": "11e516da8d025c20118ccf3b5255b999", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:35.327526", + "end_time_simulation": 9.979999542236328 + }, + "6f411ab0c8d35a14a48437f38e0c012b": { + "parameters_hash": "3eadd78bd967796aa1a2f6c1f48048be", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:35:38.246422", + "end_time_simulation": 9.979999542236328 + }, + "705a8c3ee9ef5b12a996116646911d5f": { + "parameters_hash": "ef1b25f6ce46fbdb93f7a6c884a6c632", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:40.879973", + "end_time_simulation": 9.979999542236328 + }, + "a310c7a985aa5fe0b2ca4983cb5eaf8a": { + "parameters_hash": "303e40a370e827cb5f37f0f1169373c1", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:35:44.003779", + "end_time_simulation": 9.979999542236328 + }, + "4a93f2017b7a58c8a1beaf142cc3ae1b": { + "parameters_hash": "c8d26904b8ae8f4d999b3f2e1710972a", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:46.429269", + "end_time_simulation": 9.979999542236328 + }, + "9353dc40eb7859d598a5d31596ede7ac": { + "parameters_hash": "64bec567731a3ce229de6618c6487436", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:35:49.765579", + "end_time_simulation": 9.979999542236328 + }, + "75c12b79c29655a69bd32787881f0f40": { + "parameters_hash": "e6e912edd2317cab8600c0d8ff701018", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:52.617381", + "end_time_simulation": 9.979999542236328 + }, + "d176f6081f5854db93ea218fd36b7dc1": { + "parameters_hash": "e7eaae155da190b0996d2a30369b48f5", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:35:55.247296", + "end_time_simulation": 9.979999542236328 + }, + "6040fbf0dd3552af8a12e5394be57a0e": { + "parameters_hash": "f70259f9bd03236340d3f440718ecfbc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:35:57.421238", + "end_time_simulation": 9.979999542236328 + }, + "3273b1d94d045773a3ed5d180a4511dc": { + "parameters_hash": "cdd906c5b95a94a735bd54c97bd58a2a", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:36:00.601361", + "end_time_simulation": 9.979999542236328 + }, + "cd0765daa4f254f680b9f67bade66cd9": { + "parameters_hash": "e84755ce28e4a21d7a5bcddfbc708a01", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:02.955090", + "end_time_simulation": 9.979999542236328 + }, + "22eb60ad8f495387903e93a80cf12294": { + "parameters_hash": "c888da4a3730985803ba2de913f25dad", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:36:05.366742", + "end_time_simulation": 9.979999542236328 + }, + "9646ae5511015044b647887e96eb8dca": { + "parameters_hash": "85c972828adf0db2de946c8143aa3955", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:36:08.209169", + "end_time_simulation": 9.979999542236328 + }, + "0032ee7233d45878a377fdb6063442c7": { + "parameters_hash": "b603fedf646fee16f54056e5d2600f90", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:11.020900", + "end_time_simulation": 9.979999542236328 + }, + "8515beff973352aea7ad964579bb9044": { + "parameters_hash": "0a954b15202e34084de39284ff9bd6c7", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:36:14.198883", + "end_time_simulation": 9.979999542236328 + }, + "09656866c166571783a2fbf36087c148": { + "parameters_hash": "581a5935c17964e5b2f394cf3f28a157", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:16.870128", + "end_time_simulation": 9.979999542236328 + }, + "902dbbac5eaf504581af2273eb90e81c": { + "parameters_hash": "555f18352f9b17116a2183646d64eeb1", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:36:20.108177", + "end_time_simulation": 9.979999542236328 + }, + "22d57ee61c6458b29eb06df93505bfcb": { + "parameters_hash": "54f0ad160a654f6a78f03f0648b568a6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:22.753554", + "end_time_simulation": 9.979999542236328 + }, + "70d16c10a3ed599bb6addc365d2b3db9": { + "parameters_hash": "796eede7a94d14fe65df2ec05a9a40b4", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:36:26.086524", + "end_time_simulation": 9.979999542236328 + }, + "e4ae711b5dee5f778967d897b5fd60a3": { + "parameters_hash": "300eac55a7f14518568fe5e27bb77886", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:28.794234", + "end_time_simulation": 9.979999542236328 + }, + "5a4227b6833c5aaab99cefa058084b6e": { + "parameters_hash": "0dc7f6a0b518c501d5bb55630c634a49", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:36:31.790113", + "end_time_simulation": 9.979999542236328 + }, + "5e3d703f5ba9508981369c7d341ea009": { + "parameters_hash": "1e8cf87f2e6006becdd336885ab58a02", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:34.019124", + "end_time_simulation": 9.979999542236328 + }, + "08cc400df5ba5ce08d26f495ea00c718": { + "parameters_hash": "31f4cddb001c103752d95c863af809fb", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:36:36.592754", + "end_time_simulation": 9.979999542236328 + }, + "edcab30fb4425f7d822893e3ae8ddc0c": { + "parameters_hash": "e36e013a1ea9e7efb14a0a2d2ebcd9ee", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:39.387856", + "end_time_simulation": 9.979999542236328 + }, + "e4156c6755bc5b86ab4fb7c158deb515": { + "parameters_hash": "4cb15dd030ef6fa6dee844e600116d8d", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:36:41.601821", + "end_time_simulation": 9.979999542236328 + }, + "f6fb9061905f5650a9cdc2ff306b7c88": { + "parameters_hash": "5b656b2642481ce017c599834e346b9d", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:43.764796", + "end_time_simulation": 9.979999542236328 + }, + "e40e7855b79058599803ffe83e69df2f": { + "parameters_hash": "c3c397052ce82a4ff250f2f81f6d7e25", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:36:46.384308", + "end_time_simulation": 9.979999542236328 + }, + "b7a68e5e5511505db441334199e913f2": { + "parameters_hash": "e3d9487da2e825dbbff209f7bec29429", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:49.667898", + "end_time_simulation": 9.979999542236328 + }, + "03e29327aace5e31b013ba072cc4a3a9": { + "parameters_hash": "cd1ec3dde26f86d6ef6f8bba95393def", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:36:52.170447", + "end_time_simulation": 9.979999542236328 + }, + "ba262d647be45abaa0e5bb52a9166ed9": { + "parameters_hash": "d7f49a8d12dab7a9227bd870c69535d4", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:36:54.589143", + "end_time_simulation": 9.979999542236328 + }, + "cca08f9bbb0a5d88afe4e492ce2211c9": { + "parameters_hash": "73214289f77410d376d5fd1bbc268e79", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:36:57.179097", + "end_time_simulation": 9.979999542236328 + }, + "effe796e83b45cceadb025fba657edee": { + "parameters_hash": "f7583a1e0d9cd216bc4d3b659e8dafa1", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:00.058683", + "end_time_simulation": 9.979999542236328 + }, + "f3198e91ff7d593b8b1d00e99b154708": { + "parameters_hash": "318e2ab6ec33d6dc88af05059f1c205b", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:37:03.162517", + "end_time_simulation": 9.979999542236328 + }, + "d2354df65fff5ddf9e1b8e5d138282bc": { + "parameters_hash": "0d21d8f3c8d4e08b0762b8951b3374b3", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:37:06.371435", + "end_time_simulation": 9.979999542236328 + }, + "d101fc6766415572b7fab6de6408a4dd": { + "parameters_hash": "4e6dec5d272bc169500dae78c64a06f0", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:09.535748", + "end_time_simulation": 9.979999542236328 + }, + "23e984e0745657b98c44670640edcf6f": { + "parameters_hash": "7f8608a5ffea9e5862c6d454e1218bf7", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:37:12.368099", + "end_time_simulation": 9.979999542236328 + }, + "0e3875de6d775a83a09d216c5d6d9e57": { + "parameters_hash": "f35abc0be16c509c101ac6d68b8ae297", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:15.251750", + "end_time_simulation": 9.979999542236328 + }, + "bb5c824b74c65fd0bfdeeb1b5969e46c": { + "parameters_hash": "1041e14f858ba45a53f9382d423a1bff", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:37:17.998102", + "end_time_simulation": 9.979999542236328 + }, + "4beb840fc39c517083c0580812690494": { + "parameters_hash": "127454b224c3e53f19b5e0d5dcb4c19b", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:20.267741", + "end_time_simulation": 9.979999542236328 + }, + "d53df53fc110513fab0cf1f28bdc33fb": { + "parameters_hash": "d3f0defdc2156de4327193353a40f1ac", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:37:23.675711", + "end_time_simulation": 9.979999542236328 + }, + "72fe67cc09415dfb9deb0c4e21805c89": { + "parameters_hash": "400f68ead9cb631d7fb7905aee1784ec", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:26.363197", + "end_time_simulation": 9.979999542236328 + }, + "0b0cfba8053e5f45a30e6d52ce96692d": { + "parameters_hash": "357b09fdac241f53347664760b23dc55", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:37:29.501922", + "end_time_simulation": 9.979999542236328 + }, + "0d6f17dd1b6c5e9995c0bd46408842fd": { + "parameters_hash": "21a55c1e35141304b16566e7b5f34853", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:32.257549", + "end_time_simulation": 9.979999542236328 + }, + "26c1c365fb325032b102c7b48ef670cf": { + "parameters_hash": "c62e276a2f7531fc697924d021d19fd9", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:37:35.014188", + "end_time_simulation": 9.979999542236328 + }, + "cd45d7d2aa275b179dfeebb9967a3bc2": { + "parameters_hash": "87a5200ebcca8734cbcf322e6d728b2c", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:37.407866", + "end_time_simulation": 9.979999542236328 + }, + "e2cc7f4083515d948057472541e749f9": { + "parameters_hash": "7cd0b94e84b5ae80e347e423efa65420", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:37:40.283493", + "end_time_simulation": 9.979999542236328 + }, + "a0ff785f100457e3ba3ed648b9204ef3": { + "parameters_hash": "a9d5edb3d3283ef3519c3b26cc693bbc", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:43.593979", + "end_time_simulation": 9.979999542236328 + }, + "ead1bd72fabc53758d90a08e1dcd2507": { + "parameters_hash": "98cdbe182bdd85da4d92af494499240c", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:37:46.685187", + "end_time_simulation": 9.979999542236328 + }, + "ef16dabeeb395b6899d912f0a586e298": { + "parameters_hash": "3a9ffd1c91d47b1258c237659d0c63c7", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:49.119909", + "end_time_simulation": 9.979999542236328 + }, + "95cb0d86eab35b668227df0389ec4352": { + "parameters_hash": "6b3709a6ab7044b42eeb72d2dc452dc5", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:37:52.216235", + "end_time_simulation": 9.979999542236328 + }, + "43865efdfb8c55fe8b5574f2d1b83b1f": { + "parameters_hash": "7790798d3fc1c72ba239b81f35160eea", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:54.616780", + "end_time_simulation": 9.979999542236328 + }, + "bc559b6f1f8e5336bfd2860509648fd3": { + "parameters_hash": "b318adc84b64e8b38626bc9e75c9074a", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:37:56.833336", + "end_time_simulation": 9.979999542236328 + }, + "5568926caf765b4086d61fae0ed2bc93": { + "parameters_hash": "bf9693067b4070a463dc18a0c91f90e3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:37:59.119099", + "end_time_simulation": 9.979999542236328 + }, + "d8108f5dafb05591a0165983da77f78c": { + "parameters_hash": "8828271b94284dc466380acbbabf6e83", + "run_type": "baseline", + "class_internal": "no_parameter_change", + "class_agent_facing_name": "no parameter changed", + "status": "success", + "timestamp": "2026-05-04T17:38:01.755601", + "end_time_simulation": 9.979999542236328 + }, + "acf308828c13580dbb037451e96a2244": { + "parameters_hash": "9966bbd5478730348d61cd7a7b2f046e", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:38:04.660655", + "end_time_simulation": 9.979999542236328 + }, + "6b08c67fdb425d429e2d369b71dda8dd": { + "parameters_hash": "ceea405688c73dae417bbabfc9f257f3", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:07.396984", + "end_time_simulation": 9.979999542236328 + }, + "1a4bb3d586465880a0be772d9bf056f4": { + "parameters_hash": "8a469ac6845e345f2d3342d76dc2e39e", + "run_type": "intervention", + "class_internal": "gravity_constant", + "class_agent_facing_name": "gravity_constant", + "status": "success", + "timestamp": "2026-05-04T17:38:10.326199", + "end_time_simulation": 9.979999542236328 + }, + "128dd8b963ef5a17b165abad87e70e90": { + "parameters_hash": "253b5c3d54b053139854062894fcabec", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:12.981232", + "end_time_simulation": 9.979999542236328 + }, + "825694c3733f554c88f56fb84d621eaf": { + "parameters_hash": "ba6e61a994c4b8a5dd69cbd12ae5e404", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:38:15.793189", + "end_time_simulation": 9.979999542236328 + }, + "be61d32fa437582cb44d4c90ef0ad687": { + "parameters_hash": "fe7227623614ad780f6141b4739ffe76", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:19.556781", + "end_time_simulation": 9.979999542236328 + }, + "e459f3032a4152a496af1745a8b75c10": { + "parameters_hash": "570396e18c3280218b96bfaf41280dc0", + "run_type": "intervention", + "class_internal": "plane_inclination", + "class_agent_facing_name": "plane_inclination", + "status": "success", + "timestamp": "2026-05-04T17:38:23.114868", + "end_time_simulation": 9.979999542236328 + }, + "a3be207bbbb359caa0bdc51dd6440085": { + "parameters_hash": "0d04150d47020a5a62b592c386491db6", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:25.493727", + "end_time_simulation": 9.979999542236328 + }, + "a751ca87e22257f5a98cb9cbb5f5c6ba": { + "parameters_hash": "3fd4a2eecff453b4eddf6545d7583819", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:38:28.081919", + "end_time_simulation": 9.979999542236328 + }, + "bf80ac9593cd5182ac6247d4b4027f25": { + "parameters_hash": "a48b2415f24fc8796e74ae59e0d08557", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:30.994483", + "end_time_simulation": 9.979999542236328 + }, + "67da46663e825982a82d3d439c78db7f": { + "parameters_hash": "491ce7b6ac8f29abc5d2b5949e2423bc", + "run_type": "intervention", + "class_internal": "mass", + "class_agent_facing_name": "mass", + "status": "success", + "timestamp": "2026-05-04T17:38:33.762654", + "end_time_simulation": 9.979999542236328 + }, + "8a1552f92e2c5604bda1268c7e4b9c1b": { + "parameters_hash": "7aeefdbafa4a927ad50fc8125dd98afb", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:35.942216", + "end_time_simulation": 9.979999542236328 + }, + "cca04630bd315d5092ae4e41ef3e6ab1": { + "parameters_hash": "a42d64ddd6fb6c4317a46c413a25fe74", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:38:38.631352", + "end_time_simulation": 9.979999542236328 + }, + "d95f0f1887d95368b58aa39ede338e5a": { + "parameters_hash": "405efb94604b17fdfedd51d916f991c2", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:40.900409", + "end_time_simulation": 9.979999542236328 + }, + "1c9d24d4efc4565380f4757b00650532": { + "parameters_hash": "a46e0c3226c7ddb32128abe51de6fe3f", + "run_type": "intervention", + "class_internal": "coulomb_friction_coefficient", + "class_agent_facing_name": "coulomb_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:38:44.062145", + "end_time_simulation": 9.979999542236328 + }, + "3565a4c819fd5331b840850d4eaf9152": { + "parameters_hash": "3d9d142d498447fdb196b1bc1defd810", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:46.940634", + "end_time_simulation": 9.979999542236328 + }, + "76a094cf7412561fa86c179f24507cd8": { + "parameters_hash": "68212fb103cdae6a1f018429eaaf5e1b", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:38:49.506342", + "end_time_simulation": 9.979999542236328 + }, + "c85ffb5cad445e4aa6d7de806afde64f": { + "parameters_hash": "6b34b716d886ea4ed5c74b75f69dd4ac", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:52.276120", + "end_time_simulation": 9.979999542236328 + }, + "7e43f50044175daa901fe5c26bf84e00": { + "parameters_hash": "354b1c9cd7b00e63020508947c99b967", + "run_type": "intervention", + "class_internal": "breakaway_friction_coefficient", + "class_agent_facing_name": "breakaway_friction_coefficient", + "status": "success", + "timestamp": "2026-05-04T17:38:55.384213", + "end_time_simulation": 9.979999542236328 + }, + "587c3a5460675e68b5a1466a92916732": { + "parameters_hash": "c4d45e951ff39d36c81b7662069f5d01", + "run_type": "time0_baseline", + "class_internal": "", + "class_agent_facing_name": "", + "status": "success", + "timestamp": "2026-05-04T17:38:57.796213", + "end_time_simulation": 9.979999542236328 + } +} \ No newline at end of file diff --git a/questions/MassSlide/noise_adder.py b/questions/MassSlide/noise_adder.py new file mode 100644 index 0000000000000000000000000000000000000000..6a4dee6142518a5c460b443c491a2552709d6929 --- /dev/null +++ b/questions/MassSlide/noise_adder.py @@ -0,0 +1,234 @@ +from __future__ import annotations + +import numpy as np +import pandas as pd + +from shared.noise_analysis import ( + DOCUMENTED_LOCAL_NOISE_ANALYSIS_POST_ROWS, + DOCUMENTED_LOCAL_NOISE_ANALYSIS_PRE_ROWS, + first_detectable_time_from_baseline, + quantify_analysis, +) +from shared.noise_snr import hash_string + + +_VELOCITY_COLUMN = "mass_velocity" +_FRICTION_COLUMN = "friction_force" +_NORMAL_COLUMN = "normal_force" +LOW = { + "velocity_base_sigma_scale": 0.003, + "velocity_hetero_sigma_scale": 0.003, + "velocity_drift_sigma_scale": 0.0005, + "velocity_event_sigma_scale": 0.01, + "friction_base_sigma_scale": 0.001, + "friction_drift_sigma_scale": 0.001, + "friction_bias_sigma_scale": 0.002, + "friction_event_sigma_scale": 0.02, + "normal_base_sigma_scale": 0.001, + "normal_drift_sigma_scale": 0.0002, +} +HIGH_SCALE = 4.0 +HIGH = {key: float(value) * HIGH_SCALE for key, value in LOW.items()} + + +NOISE_DICT = {"low": LOW, "high": HIGH} +SNR_THR_DICT = { + "low": {"global": [-1.0e9, -1.0e9, -1.0e9], "local": [-1.0e9, -1.0e9, -1.0e9]}, + "high": {"global": [-1.0e9, -1.0e9, -1.0e9], "local": [-1.0e9, -1.0e9, -1.0e9]}, +} +_MAX_NOISE_RESAMPLE_ATTEMPTS = 25 + + +def _analysis_meets_thresholds( + noise_analysis: dict[str, list[float | str | None]], + *, + noise_level: str, +) -> bool: + thresholds = SNR_THR_DICT[noise_level] + for scope in ("global", "local"): + values = noise_analysis.get(scope, []) + limit_values = thresholds.get(scope, []) + for idx, raw_value in enumerate(values): + if raw_value is None or idx >= len(limit_values): + continue + value = float(raw_value) + if np.isfinite(value) and value < float(limit_values[idx]): + return False + return True + + +def _rng(seed: int, key: str) -> np.random.Generator: + derived = (int(seed) ^ hash_string(key)) & 0xFFFFFFFF + return np.random.default_rng(derived) + + +def _values(df: pd.DataFrame, column: str) -> np.ndarray: + return pd.to_numeric(df[column], errors="coerce").to_numpy(dtype=float) + + +def _finite_scale(values: np.ndarray) -> float: + finite = values[np.isfinite(values)] + if finite.size == 0: + return 1.0 + spread = float(np.nanmax(finite) - np.nanmin(finite)) + rms = float(np.sqrt(np.mean(finite**2))) + return max(spread, rms, 1e-6) + + +def _coefficients(profile: str) -> dict[str, float]: + normalized = str(profile or "low").strip().lower() + if normalized == "low": + return LOW + if normalized == "high": + return HIGH + raise ValueError(f"Unknown noise profile '{profile}'. Expected 'low' or 'high'.") + + +def _smooth(values: np.ndarray) -> np.ndarray: + kernel = np.array([0.2, 0.3, 0.3, 0.2], dtype=float) + return np.convolve(values, kernel, mode="same") + + +def _drift(rng: np.random.Generator, n: int, scale: float) -> np.ndarray: + if n <= 0 or scale <= 0.0: + return np.zeros(n, dtype=float) + return _smooth(_smooth(rng.normal(0.0, scale, size=n))) + + +def _transition_mask(velocity: np.ndarray, friction: np.ndarray) -> np.ndarray: + n = max(velocity.size, friction.size) + out = np.zeros(n, dtype=bool) + for source in (velocity, friction): + if source.size <= 1: + continue + finite = np.isfinite(source) + scale = _finite_scale(source) + for idx in range(1, source.size): + if not (finite[idx] and finite[idx - 1]): + continue + sign_flip = np.sign(source[idx]) != np.sign(source[idx - 1]) + jump = abs(float(source[idx] - source[idx - 1])) > 0.08 * scale + near_zero = abs(float(source[idx])) < 0.02 * scale + if sign_flip or (jump and near_zero): + lo = max(0, idx - 2) + hi = min(out.size, idx + 3) + out[lo:hi] = True + return out + + +def _add_noise_once(df: pd.DataFrame, seed: int = 0, profile: str = "low") -> pd.DataFrame: + coeffs = _coefficients(profile) + out = df.copy() + velocity = _values(out, _VELOCITY_COLUMN) if _VELOCITY_COLUMN in out.columns else np.array([], dtype=float) + friction = _values(out, _FRICTION_COLUMN) if _FRICTION_COLUMN in out.columns else np.array([], dtype=float) + transitions = _transition_mask(velocity, friction) + + if _VELOCITY_COLUMN in out.columns: + values = velocity + scale = _finite_scale(values) + rng = _rng(seed, _VELOCITY_COLUMN) + local = np.abs(values) + ref = float(np.nanmedian(local[np.isfinite(local)])) if np.isfinite(local).any() else 0.0 + noisy = values.copy() + noisy += rng.normal( + 0.0, + coeffs["velocity_base_sigma_scale"] * scale + + coeffs["velocity_hetero_sigma_scale"] * np.maximum(local, ref), + size=values.size, + ) + noisy += _drift(rng, values.size, coeffs["velocity_drift_sigma_scale"] * scale) + if transitions.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["velocity_event_sigma_scale"] * scale, + size=values.size, + ) + * transitions.astype(float) + ) + out[_VELOCITY_COLUMN] = noisy + + if _FRICTION_COLUMN in out.columns: + values = friction + scale = _finite_scale(values) + rng = _rng(seed, _FRICTION_COLUMN) + noisy = values.copy() + noisy += rng.normal( + 0.0, + coeffs["friction_base_sigma_scale"] * scale, + size=values.size, + ) + noisy += _drift(rng, values.size, coeffs["friction_drift_sigma_scale"] * scale) + noisy += np.sign(np.nan_to_num(values, nan=0.0)) * rng.normal( + 0.0, + coeffs["friction_bias_sigma_scale"] * scale, + ) + if transitions.size == values.size: + noisy += ( + rng.normal( + 0.0, + coeffs["friction_event_sigma_scale"] * scale, + size=values.size, + ) + * transitions.astype(float) + ) + out[_FRICTION_COLUMN] = noisy + + if _NORMAL_COLUMN in out.columns: + values = _values(out, _NORMAL_COLUMN) + scale = _finite_scale(values) + rng = _rng(seed, _NORMAL_COLUMN) + noisy = values.copy() + noisy += rng.normal( + 0.0, + coeffs["normal_base_sigma_scale"] * scale, + size=values.size, + ) + noisy += _drift(rng, values.size, coeffs["normal_drift_sigma_scale"] * scale) + out[_NORMAL_COLUMN] = np.maximum(noisy, 0.0) + + return out + + +def quantify_noise( + clean: pd.DataFrame, + noisy: pd.DataFrame, + baseline: pd.DataFrame | None, +) -> dict[str, list[float | str | None]]: + first_diff = first_detectable_time_from_baseline(clean, baseline) + analysis = quantify_analysis( + clean, + noisy, + reference_df=baseline, + first_diff=first_diff, + local_pre_rows=DOCUMENTED_LOCAL_NOISE_ANALYSIS_PRE_ROWS, + local_post_rows=DOCUMENTED_LOCAL_NOISE_ANALYSIS_POST_ROWS, + ) + if first_diff is None or "local" not in analysis: + analysis["local"] = [None] * len(analysis.get("global", [])) + return analysis + + +def add_noise( + clean: pd.DataFrame, + baseline: pd.DataFrame | None, + seed: int = 0, + noise_level: str = "low", +) -> tuple[pd.DataFrame, dict[str, list[float | str | None]]]: + normalized = str(noise_level or "low").strip().lower() + if normalized not in NOISE_DICT: + raise ValueError(f"Unknown noise level '{noise_level}'. Expected 'low' or 'high'.") + current_seed = int(seed) + for _attempt in range(_MAX_NOISE_RESAMPLE_ATTEMPTS + 1): + noisy_df = _add_noise_once(clean, seed=current_seed, profile=normalized) + noise_analysis = quantify_noise(clean, noisy_df, baseline) + if _analysis_meets_thresholds(noise_analysis, noise_level=normalized): + return noisy_df, noise_analysis + current_seed += 1000 + raise RuntimeError( + f"Could not satisfy minimum SNR thresholds for noise level '{normalized}' " + f"after {_MAX_NOISE_RESAMPLE_ATTEMPTS + 1} attempts." + ) + + +__all__ = ["HIGH", "LOW", "NOISE_DICT", "SNR_THR_DICT", "add_noise", "quantify_noise"] diff --git a/questions/MassSlide/questions.json b/questions/MassSlide/questions.json new file mode 100644 index 0000000000000000000000000000000000000000..424f250d609208b6dac252d8554f367ebc95ef10 --- /dev/null +++ b/questions/MassSlide/questions.json @@ -0,0 +1,29023 @@ +{ + "version": 8, + "questions": { + "frost_01234-anchor_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "frost_01234-anchor_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-anchor" + } + }, + "fern_01234-anchor_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "fern_01234-anchor_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-anchor" + } + }, + "gentle_01234-anchor_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "gentle_01234-anchor_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-anchor" + } + }, + "frost_01234-cloud_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "frost_01234-cloud_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-cloud" + } + }, + "fern_01234-cloud_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "fern_01234-cloud_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-cloud" + } + }, + "gentle_01234-cloud_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "gentle_01234-cloud_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-cloud" + } + }, + "frost_01234-pine_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "frost_01234-pine_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-pine" + } + }, + "fern_01234-pine_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "fern_01234-pine_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-pine" + } + }, + "gentle_01234-pine_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "gentle_01234-pine_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-pine" + } + }, + "frost_01234-prairie_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "frost_01234-prairie_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-prairie" + } + }, + "fern_01234-prairie_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "fern_01234-prairie_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-prairie" + } + }, + "gentle_01234-prairie_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "gentle_01234-prairie_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-prairie" + } + }, + "frost_01234-spruce_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "frost_01234-spruce_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-spruce" + } + }, + "fern_01234-spruce_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "fern_01234-spruce_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-spruce" + } + }, + "gentle_01234-spruce_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "gentle_01234-spruce_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-spruce" + } + }, + "frost_01234-comet_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "frost_01234-comet_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "The time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-comet" + } + }, + "fern_01234-comet_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "fern_01234-comet_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-comet" + } + }, + "gentle_01234-comet_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "gentle_01234-comet_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "The time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-comet" + } + }, + "frost_01234-meadow_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "frost_01234-meadow_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-meadow" + } + }, + "fern_01234-meadow_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "fern_01234-meadow_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-meadow" + } + }, + "gentle_01234-meadow_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "gentle_01234-meadow_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-meadow" + } + }, + "frost_01234-river_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "frost_01234-river_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-river" + } + }, + "fern_01234-river_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "fern_01234-river_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-river" + } + }, + "gentle_01234-river_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "gentle_01234-river_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-river" + } + }, + "frost_01234-harbor_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "frost_01234-harbor_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-harbor" + } + }, + "fern_01234-harbor_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "fern_01234-harbor_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-harbor" + } + }, + "gentle_01234-harbor_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "gentle_01234-harbor_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled `train_samples/` directory. The corresponding labels are available in `train_labels.json` file.", + "task_artifact": "Task:\nCreate a file named results.json in the current working directory.", + "prediction_format": "For each file in test_samples/, return a ranked list of labels.\nThe first label is your final top-1 prediction and should be the single label you think is most likely correct.\nYou may include additional labels only when the evidence is genuinely ambiguous.\nAdditional labels are treated as lower-confidence alternatives.\nThe output must be valid JSON with exactly this structure:\n{\n\n \"\": [\"\"],\n \"\": [\"\", \"\"]\n\n}", + "mode_specific_requirements": "Requirements:\n- Include one entry for every Parquet file in test_samples/.\n- Every returned label must exactly match one of the allowed labels.\n- The order of labels matters: the first label is the top-1 prediction.\n- Do not include duplicate labels for a sample.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label in each returned list is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "direct", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-harbor" + } + }, + "frost_01234-willow_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "frost_01234-willow_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-willow" + } + }, + "fern_01234-willow_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "fern_01234-willow_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-willow" + } + }, + "gentle_01234-willow_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "gentle_01234-willow_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-willow" + } + }, + "frost_01234-flame_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "frost_01234-flame_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-flame" + } + }, + "fern_01234-flame_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "fern_01234-flame_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-flame" + } + }, + "gentle_01234-flame_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "gentle_01234-flame_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-flame" + } + }, + "frost_01234-orbit_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "frost_01234-orbit_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-orbit" + } + }, + "fern_01234-orbit_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "fern_01234-orbit_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-orbit" + } + }, + "gentle_01234-orbit_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "gentle_01234-orbit_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-orbit" + } + }, + "frost_01234-trail_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "frost_01234-trail_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-trail" + } + }, + "fern_01234-trail_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "fern_01234-trail_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-trail" + } + }, + "gentle_01234-trail_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "gentle_01234-trail_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-trail" + } + }, + "frost_01234-island_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "frost_01234-island_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-island" + } + }, + "fern_01234-island_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "fern_01234-island_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-island" + } + }, + "gentle_01234-island_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "gentle_01234-island_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-island" + } + }, + "frost_01234-glade_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "frost_01234-glade_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-glade" + } + }, + "fern_01234-glade_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "fern_01234-glade_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-glade" + } + }, + "gentle_01234-glade_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "gentle_01234-glade_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-glade" + } + }, + "frost_01234-canyon_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "frost_01234-canyon_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-canyon" + } + }, + "fern_01234-canyon_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "fern_01234-canyon_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-canyon" + } + }, + "gentle_01234-canyon_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "gentle_01234-canyon_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-canyon" + } + }, + "frost_01234-ember_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "frost_01234-ember_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-ember" + } + }, + "fern_01234-ember_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "fern_01234-ember_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-ember" + } + }, + "gentle_01234-ember_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "gentle_01234-ember_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-ember" + } + }, + "frost_01234-tide_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "frost_01234-tide_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-tide" + } + }, + "fern_01234-tide_0": { + "question_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_1": { + "question_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_2": { + "question_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_3": { + "question_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "fern_01234-tide_4": { + "question_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 1, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "fern_01234", + "row_slug": "fern_01234-tide" + } + }, + "gentle_01234-tide_0": { + "question_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_1": { + "question_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_2": { + "question_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_3": { + "question_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "gentle_01234-tide_4": { + "question_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ and train_samples/ folders were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "To help with this task, you can use the labeled train_samples/ directory while developing rule.py. The corresponding labels are available in train_labels.json.", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect train_samples/ and train_labels.json while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including train_samples/, test_samples/ or train_labels.json\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "code", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 3, + "number_test_samples": 10, + "is_adversarial": true, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "gentle_01234", + "row_slug": "gentle_01234-tide" + } + }, + "frost_01234-crest_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-crest_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-crest" + } + }, + "frost_01234-star_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-star_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-star" + } + }, + "frost_01234-forest_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-forest_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a block of mass m moving along an infinitely long rigid plane inclined at an angle theta under gravitational acceleration g and Coulomb friction.\nThe coordinate axis is aligned with the plane. \nThe block is also subject to an externally applied periodic force along the plane.\nThe friction force acts along the plane and opposes motion.\nWhen the block is moving, that is, when v(t) is nonzero, friction is modeled as kinetic Coulomb friction.\nA breakaway static-friction threshold is also modeled: when the block is at rest, motion starts only if the net driving force along the plane exceeds a breakaway limit.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "high", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-forest" + } + }, + "frost_01234-lagoon_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-lagoon_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-lagoon" + } + }, + "frost_01234-quartz_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-quartz_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-quartz" + } + }, + "frost_01234-dawn_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-dawn_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated", + "environment_description": "by simulating a physical process.", + "observed_columns": "Observed Signals:\ncol1: velocity of the mass along the plane\ncol2: friction force\ncol3: normal force\ncol4: time", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter among the allowed labels changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "Coulomb friction coefficient", + "breakaway friction coefficient", + "gravity acceleration", + "plane inclination angle", + "no parameter change" + ], + "label_space": "Allowed labels:\n[\"Coulomb friction coefficient\", \"breakaway friction coefficient\", \"gravity acceleration\", \"plane inclination angle\", \"no parameter change\"]", + "no_change_guidance": "Use \"no parameter change\" if there is no evidence in the data of a parameter change during the observed interval.", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + " " + ], + [ + "environment_description", + "\n\n" + ], + [ + "observed_columns", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "no_change_guidance", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "low", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-dawn" + } + }, + "frost_01234-summit_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-summit_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "none", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-summit" + } + }, + "frost_01234-aurora_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-aurora_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "low", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-aurora" + } + }, + "frost_01234-valley_0": { + "question_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 0, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_1": { + "question_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 1, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_2": { + "question_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 2, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_3": { + "question_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 3, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + }, + "frost_01234-valley_4": { + "question_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "question_text": { + "sample_source": "Context:\nThe time series in the test_samples/ folder were generated by a simulator of an unknown physical phenomenon.", + "environment_description": "The column meanings are unknown, except for the last column, which represents time.", + "observed_columns": "", + "intervention_semantics": "For each simulation, either no parameter changes occur, or exactly one parameter corresponding to one of the allowed changes during the observed simulation interval.\nIf a parameter changes, it undergoes a single instantaneous step change at an unknown time during the observed interval.", + "allowed_labels": [ + "label_0", + "label_1", + "label_2", + "label_3", + "label_4" + ], + "label_space": "[\"label_0\", \"label_1\", \"label_2\", \"label_3\"] denote different parameter changes, while \"label_4\" denotes that no parameter changed.", + "no_change_guidance": "", + "fewshot_context": "", + "task_artifact": "Task:\nCreate a Python script named rule.py in the current working directory.\nThe script must define exactly this function:\ndef predict(df) -> list[str]:", + "prediction_format": "The input df is a pandas DataFrame containing one sample with columns col1, col2, col3, and col4.\nThe function will be called on samples inside test_samples/ and on additional held-out samples with the same schema and label set.\nFor each dataframe, predict(df) must return a ranked list of labels. The first label is the final top-1 prediction and should be the single label most likely to be correct. Additional labels are optional lower-confidence alternatives and should only be included when the evidence is genuinely ambiguous.", + "mode_specific_requirements": "Requirements for predict(df):\n- Return a Python list of strings.\n- Every returned label must exactly match one of the allowed labels.\n- The first returned label is the top-1 prediction.\n- Do not include duplicate labels.\n- Do not return all labels unless the evidence is genuinely ambiguous across all labels.\n- You may inspect any file while developing rule.py, but the final submitted rule.py must not read, open, import, or depend on any files at prediction time, including test_samples/\n- The final rule.py must be able to run on a dataframe alone.", + "evaluation": "Evaluation:\nThe primary evaluation metric is top-1 accuracy: the first label returned by predict(df) is compared with the hidden correct label.\nA secondary shortlist score may also be reported. For a returned list of length m, the sample receives score 1/m if the hidden correct label appears anywhere in the list, and 0 otherwise. Therefore, unnecessary extra labels reduce the secondary score.", + "runtime_constraints": "Additional requirements:\n- If you create intermediate files, images, scripts, or notes while solving the task, create them in the current working directory.\n- Internet access is disabled.", + "ordered_field_agent_prompt": [ + [ + "sample_source", + "\n" + ], + [ + "environment_description", + "\n\n" + ], + [ + "intervention_semantics", + "\n\n" + ], + [ + "label_space", + "\n\n" + ], + [ + "task_artifact", + "\n" + ], + [ + "prediction_format", + "\n\n" + ], + [ + "fewshot_context", + "\n\n" + ], + [ + "mode_specific_requirements", + "\n\n" + ], + [ + "evaluation", + "\n\n" + ], + [ + "runtime_constraints", + "" + ] + ] + }, + "recipe_info": { + "type_of_request": "open-ended", + "desc_level": "none", + "noise_level": "high", + "number_train_samples_per_class": 0, + "number_test_samples": 10, + "is_adversarial": null, + "question_seed": 4, + "test_set_slug": "home", + "shot_slug": "frost_01234", + "row_slug": "frost_01234-valley" + } + } + }, + "label_int_mapping": { + "Coulomb friction coefficient": 0, + "breakaway friction coefficient": 1, + "gravity acceleration": 2, + "plane inclination angle": 3, + "no parameter change": 4 + }, + "ground_truth_information": { + "pattern_observed": {}, + "interventions": { + "c1e94ce33230421a847d36495497409d": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 0.5, + "coulomb_friction_coefficient": 0.20617170169264934, + "breakaway_friction_coefficient": 6, + "initial_velocity": 1.4812812244792246, + "Amplitude": -50.838477, + "Period": 1.900156 + }, + "intervention_time": 3.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "ca2826a9ea444e4fbc8b2cd785dfc4a3": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 0.5, + "coulomb_friction_coefficient": 0.20617170169264934, + "breakaway_friction_coefficient": 6, + "initial_velocity": 1.4812812244792246, + "Amplitude": -50.838477, + "Period": 1.900156 + }, + "intervention_time": 3.0, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.008890186332394643, + "first_diff": [ + 3.2799999713897705, + 3.0199999809265137, + null + ] + }, + "4b91015695f94331ba68ab2f9bc8feb1": { + "initial_parameters": { + "gravity_constant": 9.81, + "plane_inclination": 25, + "mass": 2.5, + "coulomb_friction_coefficient": 1.5, + "breakaway_friction_coefficient": 6, + "initial_velocity": 6, + "Amplitude": -59.16469, + "Period": 1.613908 + }, + "intervention_time": 2.0, + "changed_parameter": "gravity_constant", + "new_value": 3.1, + "first_diff": [ + 3.240000009536743, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "337efe3f652b4c12b24b70a63f25adff": { + "initial_parameters": { + "gravity_constant": 9.81, + "plane_inclination": 25, + "mass": 2.5, + "coulomb_friction_coefficient": 1.5, + "breakaway_friction_coefficient": 6, + "initial_velocity": 6, + "Amplitude": -59.16469, + "Period": 1.613908 + }, + "intervention_time": 2.0, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 3, + "first_diff": [ + 3.240000009536743, + 3.240000009536743, + null + ] + }, + "322aa53bc0ae43c08d03196c3fa01dcf": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -5, + "Amplitude": 33.57681, + "Period": 1.800831 + }, + "intervention_time": 2.5, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "2eaaaa5d6f8e4c1b8447b4389e52853e": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -5, + "Amplitude": 33.57681, + "Period": 1.800831 + }, + "intervention_time": 2.5, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 5.64293765401431, + "first_diff": [ + 3.619999885559082, + 3.619999885559082, + null + ] + }, + "825bb775cf154fdb983653ce20d329dc": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -5, + "Amplitude": 33.57681, + "Period": 1.800831 + }, + "intervention_time": 2.5, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 11.936912738420265, + "first_diff": [ + 3.619999885559082, + 3.619999885559082, + null + ] + }, + "dd028d3ab7fb4aae84694d21ecf4c201": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 4.396466461048128, + "Amplitude": -120.961652, + "Period": 1.810527 + }, + "intervention_time": 3.6, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "bdf601681b8541659ad53f2b52d11526": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3, + "Amplitude": 99.119345, + "Period": 0.768386 + }, + "intervention_time": 2.5839646968555012, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "ab1dbdc0c2344be7a4b105dd445c3e62": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3, + "Amplitude": 99.119345, + "Period": 0.768386 + }, + "intervention_time": 2.5839646968555012, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + 2.619999885559082, + 2.619999885559082, + 2.619999885559082 + ] + }, + "97f285a3809940f68129ed3a9d1d9b26": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3, + "Amplitude": 99.119345, + "Period": 0.768386 + }, + "intervention_time": 2.5839646968555012, + "changed_parameter": "plane_inclination", + "new_value": 45, + "first_diff": [ + 2.619999885559082, + 2.619999885559082, + 2.619999885559082 + ] + }, + "97c922e470a6420aaa0db38a7b47ab14": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3, + "Amplitude": 99.119345, + "Period": 0.768386 + }, + "intervention_time": 2.5839646968555012, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.2, + "first_diff": [ + 2.619999885559082, + 2.619999885559082, + null + ] + }, + "2f00e9a90e3c4484aa48107c021745aa": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 12, + "Amplitude": -77.052202, + "Period": 2.317339 + }, + "intervention_time": 2.0, + "changed_parameter": "gravity_constant", + "new_value": 2.1314429713336716, + "first_diff": [ + 2.319999933242798, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "03cf9a5fd2af43aa8b2413596e24ffa4": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 12, + "Amplitude": -77.052202, + "Period": 2.317339 + }, + "intervention_time": 6.0, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 3.1, + "first_diff": [ + 6.960000038146973, + 6.960000038146973, + null + ] + }, + "924138e0d0f9479f8dccc89cc12e589c": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 65.93838359779784, + "mass": 10, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 5.095445115010333, + "initial_velocity": 3.7412559900975477, + "Amplitude": 0.01, + "Period": 0.85302 + }, + "intervention_time": 3.409, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "3c9308c66d644b07be5b2c3ee6ffa4e6": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 65.93838359779784, + "mass": 10, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 5.095445115010333, + "initial_velocity": 3.7412559900975477, + "Amplitude": 0.01, + "Period": 0.85302 + }, + "intervention_time": 3.409, + "changed_parameter": "gravity_constant", + "new_value": 4.328057954612467, + "first_diff": [ + null, + 3.440000057220459, + 3.440000057220459 + ] + }, + "9e10b79464c24f36a53f252be6541d5f": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 65.93838359779784, + "mass": 10, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 5.095445115010333, + "initial_velocity": 3.7412559900975477, + "Amplitude": 0.01, + "Period": 0.85302 + }, + "intervention_time": 3.409, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 1.7745966692414834, + "first_diff": [ + 3.440000057220459, + 3.440000057220459, + null + ] + }, + "afec7fdef34d4256a0060406c9b1cd7c": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 10, + "Amplitude": 172.886023, + "Period": 1.737068 + }, + "intervention_time": 2.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "260750f9d27a4578ba899fcfb02ac653": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 5, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 10, + "Amplitude": 172.886023, + "Period": 1.737068 + }, + "intervention_time": 2.0, + "changed_parameter": "plane_inclination", + "new_value": 45, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "5c6ece223a1a46f6a1987449154ff138": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 35.94974796809461, + "mass": 10, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 12, + "Amplitude": 188.656209, + "Period": 1.139029 + }, + "intervention_time": 2.0, + "changed_parameter": "plane_inclination", + "new_value": 18.912609692389978, + "first_diff": [ + null, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "3d97964fb7c24cc291fbc25dea19c9d4": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 9.5, + "Amplitude": -152.443198, + "Period": 2.196553 + }, + "intervention_time": 2.0, + "changed_parameter": "gravity_constant", + "new_value": 4.314963801640732, + "first_diff": [ + null, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "cd5b8bdef4d44b7d8f6682e3cde5fcf2": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 9.5, + "Amplitude": -152.443198, + "Period": 2.196553 + }, + "intervention_time": 2.0, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 3, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + null + ] + }, + "1d716244585943cc9e58b841706c0a36": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 9.5, + "Amplitude": -152.443198, + "Period": 2.196553 + }, + "intervention_time": 2.0, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.0050158312405462644, + "first_diff": [ + 3.059999942779541, + 2.0199999809265137, + null + ] + }, + "96814fa562ac4bac8cdd1ace014be742": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -4.989035388416348, + "Amplitude": 72.419311, + "Period": 0.687527 + }, + "intervention_time": 2.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "e430606893ee431087b189d71cd29ac5": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -4.989035388416348, + "Amplitude": 72.419311, + "Period": 0.687527 + }, + "intervention_time": 2.0, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "9b572270b12942a390d67ff9595a2cbd": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -4.989035388416348, + "Amplitude": 72.419311, + "Period": 0.687527 + }, + "intervention_time": 2.0, + "changed_parameter": "plane_inclination", + "new_value": 45, + "first_diff": [ + 2.0999999046325684, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "badd89b9580a4d319ec3f8a0bee814b1": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -4.989035388416348, + "Amplitude": 72.419311, + "Period": 0.687527 + }, + "intervention_time": 2.0, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.2, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + null + ] + }, + "b6e5860643684682b60e6c146121b2be": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 7.488140773015102, + "Amplitude": 23.004466, + "Period": 0.871447 + }, + "intervention_time": 2.8, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "833168b23c474c3f95c1ba4d9fff6176": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 7.488140773015102, + "Amplitude": 23.004466, + "Period": 0.871447 + }, + "intervention_time": 2.8, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + 4.360000133514404, + 2.819999933242798, + 2.819999933242798 + ] + }, + "bd1db72c3cdb481caa778ac3639cccb3": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 7.488140773015102, + "Amplitude": 23.004466, + "Period": 0.871447 + }, + "intervention_time": 2.8, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 5.967407316575695, + "first_diff": [ + 4.360000133514404, + 3.5, + null + ] + }, + "9829e3384fae4a77ae3bc21c6c8a6aea": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 7.488140773015102, + "Amplitude": 23.004466, + "Period": 0.871447 + }, + "intervention_time": 2.8, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 10.171342832265925, + "first_diff": [ + 4.360000133514404, + 3.5, + null + ] + }, + "2713905d0dfa4d56aac01b96531ab3e7": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3.3799825244113872, + "Amplitude": -24.129366, + "Period": 1.15969 + }, + "intervention_time": 4.4, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "223c934b74a24e5f9df19997035eef2f": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3.3799825244113872, + "Amplitude": -24.129366, + "Period": 1.15969 + }, + "intervention_time": 4.4, + "changed_parameter": "plane_inclination", + "new_value": 45, + "first_diff": [ + 4.420000076293945, + 4.420000076293945, + 4.420000076293945 + ] + }, + "a788bcbddb3e4143afc049ae6fd613ed": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3.3799825244113872, + "Amplitude": -24.129366, + "Period": 1.15969 + }, + "intervention_time": 4.4, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.7745966692414834, + "first_diff": [ + 4.420000076293945, + 4.420000076293945, + null + ] + }, + "7b689cd67c2d4021930e364b1c2ed8ca": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": -3.3799825244113872, + "Amplitude": -24.129366, + "Period": 1.15969 + }, + "intervention_time": 4.4, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.032792814029319314, + "first_diff": [ + 4.5, + 4.420000076293945, + null + ] + }, + "e4faa5a0e1ad4ca4b94bf2f18f4a093f": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 11.2, + "Amplitude": -29.706859, + "Period": 2.087496 + }, + "intervention_time": 2.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "d1141599d62d4cd8a57708873e904583": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 11.2, + "Amplitude": -29.706859, + "Period": 2.087496 + }, + "intervention_time": 2.0, + "changed_parameter": "plane_inclination", + "new_value": 45, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "b6a00423859b48e8b634aaf7a4b4208a": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 11.2, + "Amplitude": -29.706859, + "Period": 2.087496 + }, + "intervention_time": 2.0, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + 2.0999999046325684, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "208ef509e8b04166ae5333a2877f1295": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 11.2, + "Amplitude": -29.706859, + "Period": 2.087496 + }, + "intervention_time": 2.0, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.2, + "first_diff": [ + 2.0999999046325684, + 2.0999999046325684, + null + ] + }, + "450660a3c2d24f5bae3eeec80638199c": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.8, + "Amplitude": -54.072054, + "Period": 1.454112 + }, + "intervention_time": 2.2, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + null, + 2.2200000286102295, + 2.2200000286102295 + ] + }, + "59b2dd9b949c42e7ab404612b867f051": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.8, + "Amplitude": -54.072054, + "Period": 1.454112 + }, + "intervention_time": 2.2, + "changed_parameter": "gravity_constant", + "new_value": 2.8243866777058297, + "first_diff": [ + 2.9200000762939453, + 2.2200000286102295, + 2.2200000286102295 + ] + }, + "5cff394d93ba45fc950eb525d975c078": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.8, + "Amplitude": -54.072054, + "Period": 1.454112 + }, + "intervention_time": 2.2, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 3.1, + "first_diff": [ + 2.9200000762939453, + 2.9200000762939453, + null + ] + }, + "b9f73c1ba2e14456bc1cb2701b0d61d4": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 0.6, + "Amplitude": -46.435217, + "Period": 1.222944 + }, + "intervention_time": 3.1, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "509ca665d1854fa596c901a5468197ab": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 0.6, + "Amplitude": -46.435217, + "Period": 1.222944 + }, + "intervention_time": 3.1, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + null, + 3.119999885559082, + 3.119999885559082 + ] + }, + "d913f421c0964aea99408cec015c64c1": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 0.6, + "Amplitude": -46.435217, + "Period": 1.222944 + }, + "intervention_time": 3.1, + "changed_parameter": "gravity_constant", + "new_value": 4.3051551706298365, + "first_diff": [ + null, + 3.119999885559082, + 3.119999885559082 + ] + }, + "8ffce32ebf95402db33ab77ea1fe5f12": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 0.6, + "Amplitude": -46.435217, + "Period": 1.222944 + }, + "intervention_time": 3.1, + "changed_parameter": "plane_inclination", + "new_value": 45, + "first_diff": [ + null, + 3.119999885559082, + 3.119999885559082 + ] + }, + "421875a8a7f441c4bffd546a137058b9": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 0.6, + "Amplitude": -46.435217, + "Period": 1.222944 + }, + "intervention_time": 3.1, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + null, + 3.119999885559082, + 3.119999885559082 + ] + }, + "52cb93e7e69d4abf85d6db293440f76c": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 6, + "initial_velocity": -1.4, + "Amplitude": 170.884575, + "Period": 2.341351 + }, + "intervention_time": 2.7, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "f664534ad3bb423eae4c699ad1d1ad79": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 6, + "initial_velocity": -1.4, + "Amplitude": 170.884575, + "Period": 2.341351 + }, + "intervention_time": 2.7, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + null, + 2.7200000286102295, + 2.7200000286102295 + ] + }, + "a3dc646b3f7647079bfb49c9acaa18e6": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 6, + "initial_velocity": -1.4, + "Amplitude": 170.884575, + "Period": 2.341351 + }, + "intervention_time": 2.7, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 1.0954451150103324, + "first_diff": [ + 4.699999809265137, + 4.699999809265137, + null + ] + }, + "00af86611bf0452986e018425db6cdd2": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 4, + "initial_velocity": 10.4, + "Amplitude": 300.74409, + "Period": 1.734409 + }, + "intervention_time": 2.0, + "changed_parameter": "gravity_constant", + "new_value": 3.4349931465905423, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "477fa4f8ca014f559465bdc2c2318c78": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 4, + "initial_velocity": 10.4, + "Amplitude": 300.74409, + "Period": 1.734409 + }, + "intervention_time": 2.0, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.2, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + null + ] + }, + "b9d5d04e374c47d58d820fca9d364a11": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 15, + "mass": 10, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 4, + "initial_velocity": 10.4, + "Amplitude": 300.74409, + "Period": 1.734409 + }, + "intervention_time": 6.0, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 12.618389132338475, + "first_diff": [ + 6.940000057220459, + 6.940000057220459, + null + ] + }, + "baf1e1790569451993d611a8cb903120": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 6.542451483081047, + "breakaway_friction_coefficient": 30.484867072759553, + "initial_velocity": 2.2, + "Amplitude": -130.971602, + "Period": 1.431655 + }, + "intervention_time": 2.0, + "changed_parameter": "plane_inclination", + "new_value": 64.37691840459206, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "bd084d19f2444af8a6f98ec7815d2882": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 6.542451483081047, + "breakaway_friction_coefficient": 30.484867072759553, + "initial_velocity": 2.2, + "Amplitude": -130.971602, + "Period": 1.431655 + }, + "intervention_time": 2.0, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + 2.0199999809265137, + 2.0199999809265137, + 2.0199999809265137 + ] + }, + "aec60d90f7ef4979a80c56f5c5bdeafa": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1.8, + "Amplitude": 23.915133, + "Period": 2.604873 + }, + "intervention_time": 3.4, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "e4a4fea8b64e496cb7e24f4de9f270b7": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1.8, + "Amplitude": 23.915133, + "Period": 2.604873 + }, + "intervention_time": 3.4, + "changed_parameter": "gravity_constant", + "new_value": 2.0462806416739, + "first_diff": [ + 3.4200000762939453, + 3.4200000762939453, + 3.4200000762939453 + ] + }, + "e6e67852460340c291d3b2d33978e955": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1.8, + "Amplitude": 23.915133, + "Period": 2.604873 + }, + "intervention_time": 3.4, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 1.05, + "first_diff": [ + 3.440000057220459, + 3.4200000762939453, + null + ] + }, + "dcc54ad4f3014c0cb2545e1f84a9ee38": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1.8, + "Amplitude": 23.915133, + "Period": 2.604873 + }, + "intervention_time": 3.4, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.2, + "first_diff": [ + 3.4200000762939453, + 3.4200000762939453, + null + ] + }, + "9e2271b4625541f7964a03844c658511": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1.8, + "Amplitude": 23.915133, + "Period": 2.604873 + }, + "intervention_time": 3.4, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 0.7840670304525944, + "first_diff": [ + 4.28000020980835, + 4.28000020980835, + null + ] + }, + "c56def5d135e40c290812174a4d3a066": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -0.8, + "Amplitude": -90.683362, + "Period": 2.412222 + }, + "intervention_time": 2.9, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "cfcf113e827240a6b7e7a4be86078864": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -0.8, + "Amplitude": -90.683362, + "Period": 2.412222 + }, + "intervention_time": 2.9, + "changed_parameter": "plane_inclination", + "new_value": 69.40740910737117, + "first_diff": [ + 2.9800000190734863, + 2.9200000762939453, + 2.9200000762939453 + ] + }, + "95357a192b404d9d942b699a96f3f91d": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 0.5, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -0.8, + "Amplitude": -90.683362, + "Period": 2.412222 + }, + "intervention_time": 2.9, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + 2.9600000381469727, + 2.9200000762939453, + 2.9200000762939453 + ] + }, + "38e9b9cde3d547a9924b3af04070203a": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 8.6, + "Amplitude": -27.652423, + "Period": 1.686371 + }, + "intervention_time": 3.1, + "changed_parameter": "gravity_constant", + "new_value": 4.382126277006615, + "first_diff": [ + null, + 3.119999885559082, + 3.119999885559082 + ] + }, + "6ff497703dc54edb8534f212c5b7c567": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 8.6, + "Amplitude": -27.652423, + "Period": 1.686371 + }, + "intervention_time": 3.1, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + null, + 3.119999885559082, + 3.119999885559082 + ] + }, + "633f8b94b4a64afab5c4183b67987092": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": 8.6, + "Amplitude": -27.652423, + "Period": 1.686371 + }, + "intervention_time": 3.1, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 3, + "first_diff": [ + 3.380000114440918, + 3.380000114440918, + null + ] + }, + "3ba3a08aac1441ab9f7062076ab2e050": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.4, + "Amplitude": 93.657438, + "Period": 0.844462 + }, + "intervention_time": 3.3, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + 3.380000114440918, + 3.319999933242798, + 3.319999933242798 + ] + }, + "bd222ab0a4034903923ada8b7a09816f": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.4, + "Amplitude": 93.657438, + "Period": 0.844462 + }, + "intervention_time": 3.3, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 3, + "first_diff": [ + 3.380000114440918, + 3.380000114440918, + null + ] + }, + "b1fb837092144b5cae1c2a952a2ea10c": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.4, + "Amplitude": 93.657438, + "Period": 0.844462 + }, + "intervention_time": 3.3, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 10.5798577746978, + "first_diff": [ + 3.380000114440918, + 3.380000114440918, + null + ] + }, + "e07a5edf78e844b8ba42f7b4add2cfb7": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 2.7, + "Amplitude": -156.281151, + "Period": 2.204563 + }, + "intervention_time": 2.7, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "0982071fbb684307913a380272a1ae91": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 2.7, + "Amplitude": -156.281151, + "Period": 2.204563 + }, + "intervention_time": 2.7, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + 2.7200000286102295, + 2.7200000286102295, + 2.7200000286102295 + ] + }, + "e8e9190868644fc38aa1113756e04fee": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 2.7, + "Amplitude": -156.281151, + "Period": 2.204563 + }, + "intervention_time": 2.7, + "changed_parameter": "plane_inclination", + "new_value": 68.99581788454233, + "first_diff": [ + 2.7200000286102295, + 2.7200000286102295, + 2.7200000286102295 + ] + }, + "0c5a280f084b4d1fa52dddd30269c1d4": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 2.23606797749979, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": 2.7, + "Amplitude": -156.281151, + "Period": 2.204563 + }, + "intervention_time": 5.7, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 60.18965727644844, + "first_diff": [ + 6.619999885559082, + 6.619999885559082, + null + ] + }, + "38482b8747bc4bfe88ac617356e9c6fe": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 42.3095512138648, + "initial_velocity": -1.4, + "Amplitude": -66.149609, + "Period": 2.532138 + }, + "intervention_time": 3.0, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "ff3a19d5eaad46c1aed4ca4373c26de3": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 42.3095512138648, + "initial_velocity": -1.4, + "Amplitude": -66.149609, + "Period": 2.532138 + }, + "intervention_time": 3.0, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + 3.0199999809265137, + 3.0199999809265137, + 3.0199999809265137 + ] + }, + "53d61728255b43158f043258bf085cb9": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 42.3095512138648, + "initial_velocity": -1.4, + "Amplitude": -66.149609, + "Period": 2.532138 + }, + "intervention_time": 3.0, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 3, + "first_diff": [ + 3.0199999809265137, + 3.0199999809265137, + null + ] + }, + "fb2367ca63f6468bbb95f4ddbd0d99c9": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1, + "Amplitude": 132.666024, + "Period": 1.953778 + }, + "intervention_time": 3.6, + "changed_parameter": "gravity_constant", + "new_value": 20, + "first_diff": [ + 3.9200000762939453, + 3.619999885559082, + 3.619999885559082 + ] + }, + "8840791a44ca45f2b1e72ee59093cef4": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1, + "Amplitude": 132.666024, + "Period": 1.953778 + }, + "intervention_time": 3.6, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + 3.619999885559082, + 3.619999885559082, + 3.619999885559082 + ] + }, + "1f0a940c521643948cd6a4a3f8c15067": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1, + "Amplitude": 132.666024, + "Period": 1.953778 + }, + "intervention_time": 3.6, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 1.05, + "first_diff": [ + 3.619999885559082, + 3.9200000762939453, + null + ] + }, + "3ad87b80437b43fda95e551a2efc7b87": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1, + "Amplitude": 132.666024, + "Period": 1.953778 + }, + "intervention_time": 3.6, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 6, + "first_diff": [ + 3.619999885559082, + 3.9200000762939453, + null + ] + }, + "299fe66074144eeeae74735cf42d161b": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 0.7745966692414834, + "breakaway_friction_coefficient": 1.0954451150103324, + "initial_velocity": 1, + "Amplitude": 132.666024, + "Period": 1.953778 + }, + "intervention_time": 3.6, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 0.7840670304525944, + "first_diff": [ + 3.619999885559082, + 3.619999885559082, + null + ] + }, + "115036fcb4cb40278192aeaec3c948ba": { + "initial_parameters": { + "gravity_constant": 6, + "plane_inclination": 45, + "mass": 10, + "coulomb_friction_coefficient": 2.1, + "breakaway_friction_coefficient": 20, + "initial_velocity": 6.1, + "Amplitude": 190.365955, + "Period": 1.542793 + }, + "intervention_time": 2.8, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "3dee5963901d530e9f2fde3c84a55982": { + "initial_parameters": { + "gravity_constant": 18, + "plane_inclination": 48, + "mass": 5.8, + "coulomb_friction_coefficient": 0.6, + "breakaway_friction_coefficient": 18, + "initial_velocity": 9, + "Amplitude": 280, + "Period": 0.75 + }, + "intervention_time": 2.38, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "3dc5e2e8d7f150fbaf80df25ba9c426a": { + "initial_parameters": { + "gravity_constant": 18, + "plane_inclination": 48, + "mass": 5.8, + "coulomb_friction_coefficient": 0.6, + "breakaway_friction_coefficient": 18, + "initial_velocity": 9, + "Amplitude": 280, + "Period": 0.75 + }, + "intervention_time": 2.5, + "changed_parameter": "gravity_constant", + "new_value": 7.2, + "first_diff": [ + null, + 2.5199999809265137, + 2.5199999809265137 + ] + }, + "7d1520b2be3c531ab4acbe645c9e18cd": { + "initial_parameters": { + "gravity_constant": 4, + "plane_inclination": 18, + "mass": 0.8, + "coulomb_friction_coefficient": 0.08, + "breakaway_friction_coefficient": 4.5, + "initial_velocity": -4.5, + "Amplitude": -320, + "Period": 2.75 + }, + "intervention_time": 2.56, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "402307421702585c86a2ff40516deca2": { + "initial_parameters": { + "gravity_constant": 4, + "plane_inclination": 18, + "mass": 0.8, + "coulomb_friction_coefficient": 0.08, + "breakaway_friction_coefficient": 4.5, + "initial_velocity": -4.5, + "Amplitude": -320, + "Period": 2.75 + }, + "intervention_time": 2.68, + "changed_parameter": "gravity_constant", + "new_value": 2, + "first_diff": [ + null, + 2.700000047683716, + 2.700000047683716 + ] + }, + "0d79edc07243565d8d8b5380ef81e340": { + "initial_parameters": { + "gravity_constant": 4, + "plane_inclination": 18, + "mass": 0.8, + "coulomb_friction_coefficient": 0.08, + "breakaway_friction_coefficient": 4.5, + "initial_velocity": -4.5, + "Amplitude": -320, + "Period": 2.75 + }, + "intervention_time": 3.2, + "changed_parameter": "plane_inclination", + "new_value": 5, + "first_diff": [ + 6.940000057220459, + 3.2200000286102295, + 3.2200000286102295 + ] + }, + "23cca9a4b30252e89b56e757b3b6b0f0": { + "initial_parameters": { + "gravity_constant": 14, + "plane_inclination": 40, + "mass": 3.2, + "coulomb_friction_coefficient": 2.2, + "breakaway_friction_coefficient": 25, + "initial_velocity": 11, + "Amplitude": -240, + "Period": 1.05 + }, + "intervention_time": 2.74, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "cc910fa7368c5587804f005866a22079": { + "initial_parameters": { + "gravity_constant": 14, + "plane_inclination": 40, + "mass": 3.2, + "coulomb_friction_coefficient": 2.2, + "breakaway_friction_coefficient": 25, + "initial_velocity": 11, + "Amplitude": -240, + "Period": 1.05 + }, + "intervention_time": 2.74, + "changed_parameter": "gravity_constant", + "new_value": 23.8, + "first_diff": [ + null, + 2.759999990463257, + 2.759999990463257 + ] + }, + "e0934b52a37759e3a90cd51c68f250e5": { + "initial_parameters": { + "gravity_constant": 7, + "plane_inclination": 32, + "mass": 1.3, + "coulomb_friction_coefficient": 0.35, + "breakaway_friction_coefficient": 9, + "initial_velocity": -2.5, + "Amplitude": 360, + "Period": 2.25 + }, + "intervention_time": 2.92, + "changed_parameter": "gravity_constant", + "new_value": 11.9, + "first_diff": [ + 3.1600000858306885, + 2.940000057220459, + 2.940000057220459 + ] + }, + "c5d7959f17da52dda612bcc328e0b427": { + "initial_parameters": { + "gravity_constant": 7, + "plane_inclination": 32, + "mass": 1.3, + "coulomb_friction_coefficient": 0.35, + "breakaway_friction_coefficient": 9, + "initial_velocity": -2.5, + "Amplitude": 360, + "Period": 2.25 + }, + "intervention_time": 3.04, + "changed_parameter": "gravity_constant", + "new_value": 2.8, + "first_diff": [ + 3.319999933242798, + 3.059999942779541, + 3.059999942779541 + ] + }, + "07fd7c347c7354bbb738d7765a2fea14": { + "initial_parameters": { + "gravity_constant": 7, + "plane_inclination": 32, + "mass": 1.3, + "coulomb_friction_coefficient": 0.35, + "breakaway_friction_coefficient": 9, + "initial_velocity": -2.5, + "Amplitude": 360, + "Period": 2.25 + }, + "intervention_time": 4.48, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 1.4, + "first_diff": [ + 4.71999979019165, + 4.5, + null + ] + }, + "3cdf2c46f20e55c0b081c3745fbf8c40": { + "initial_parameters": { + "gravity_constant": 7, + "plane_inclination": 32, + "mass": 1.3, + "coulomb_friction_coefficient": 0.35, + "breakaway_friction_coefficient": 9, + "initial_velocity": -2.5, + "Amplitude": 360, + "Period": 2.25 + }, + "intervention_time": 4.6, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.0875, + "first_diff": [ + 5.579999923706055, + 4.619999885559082, + null + ] + }, + "b794f7c284485e548f68445b0c123a3a": { + "initial_parameters": { + "gravity_constant": 7.461112, + "plane_inclination": 20.94491, + "mass": 2.892309, + "coulomb_friction_coefficient": 2.8, + "breakaway_friction_coefficient": 6, + "initial_velocity": 6.04945, + "Amplitude": -106.91249, + "Period": 1.423934 + }, + "intervention_time": 2.18, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "1dbde2fdb75b5ca294baa80857327b67": { + "initial_parameters": { + "gravity_constant": 7.461112, + "plane_inclination": 20.94491, + "mass": 2.892309, + "coulomb_friction_coefficient": 2.8, + "breakaway_friction_coefficient": 6, + "initial_velocity": 6.04945, + "Amplitude": -106.91249, + "Period": 1.423934 + }, + "intervention_time": 2.61, + "changed_parameter": "plane_inclination", + "new_value": 40.94491, + "first_diff": [ + 2.859999895095825, + 2.640000104904175, + 2.640000104904175 + ] + }, + "906b2f75d36d5647a8fe4ef05f19fa5d": { + "initial_parameters": { + "gravity_constant": 7.461112, + "plane_inclination": 20.94491, + "mass": 2.892309, + "coulomb_friction_coefficient": 2.8, + "breakaway_friction_coefficient": 6, + "initial_velocity": 6.04945, + "Amplitude": -106.91249, + "Period": 1.423934 + }, + "intervention_time": 3.99, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 3, + "first_diff": [ + 4.28000020980835, + 4.28000020980835, + null + ] + }, + "4936a5b652995d1cbb5987f3c7ee6866": { + "initial_parameters": { + "gravity_constant": 12.864407, + "plane_inclination": 38.583724, + "mass": 3.362231, + "coulomb_friction_coefficient": 1.4, + "breakaway_friction_coefficient": 6, + "initial_velocity": 10.192879, + "Amplitude": 158.906158, + "Period": 1.250754 + }, + "intervention_time": 2.4, + "changed_parameter": "gravity_constant", + "new_value": 5.36017, + "first_diff": [ + 2.5199999809265137, + 2.4200000762939453, + 2.4200000762939453 + ] + }, + "9b1982d86e3b5bb892c7d1270f5e8e4c": { + "initial_parameters": { + "gravity_constant": 12.864407, + "plane_inclination": 38.583724, + "mass": 3.362231, + "coulomb_friction_coefficient": 1.4, + "breakaway_friction_coefficient": 6, + "initial_velocity": 10.192879, + "Amplitude": 158.906158, + "Period": 1.250754 + }, + "intervention_time": 4.12, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 3, + "first_diff": [ + 5.019999980926514, + 5.019999980926514, + null + ] + }, + "ffcb3e01741a516091f8da7aeb689274": { + "initial_parameters": { + "gravity_constant": 7.619848, + "plane_inclination": 46.701831, + "mass": 1.279688, + "coulomb_friction_coefficient": 0.75, + "breakaway_friction_coefficient": 1.2, + "initial_velocity": -1.208072, + "Amplitude": 45.029167, + "Period": 1.488852 + }, + "intervention_time": 2.44, + "changed_parameter": "gravity_constant", + "new_value": 12.572749, + "first_diff": [ + 2.9800000190734863, + 2.4600000381469727, + 2.4600000381469727 + ] + }, + "52d91d5389a652baa478069ea293ca73": { + "initial_parameters": { + "gravity_constant": 7.619848, + "plane_inclination": 46.701831, + "mass": 1.279688, + "coulomb_friction_coefficient": 0.75, + "breakaway_friction_coefficient": 1.2, + "initial_velocity": -1.208072, + "Amplitude": 45.029167, + "Period": 1.488852 + }, + "intervention_time": 4.25, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 0.784, + "first_diff": [ + 4.28000020980835, + 4.28000020980835, + null + ] + }, + "0dc61f09ac3a57cba3e986abd378a58c": { + "initial_parameters": { + "gravity_constant": 16.913222, + "plane_inclination": 18.820694, + "mass": 1.214361, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": -4.057459, + "Amplitude": -33.145587, + "Period": 2.499777 + }, + "intervention_time": 2.57, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "d950be4ceaf855e4b99175c9b68e37f5": { + "initial_parameters": { + "gravity_constant": 16.913222, + "plane_inclination": 18.820694, + "mass": 1.214361, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": -4.057459, + "Amplitude": -33.145587, + "Period": 2.499777 + }, + "intervention_time": 3.0, + "changed_parameter": "plane_inclination", + "new_value": 38.820694, + "first_diff": [ + 3.0399999618530273, + 3.0199999809265137, + 3.0199999809265137 + ] + }, + "0655d91288195e93a7903500183734dd": { + "initial_parameters": { + "gravity_constant": 16.913222, + "plane_inclination": 18.820694, + "mass": 1.214361, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": -4.057459, + "Amplitude": -33.145587, + "Period": 2.499777 + }, + "intervention_time": 3.86, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.59, + "first_diff": [ + 3.9000000953674316, + 3.880000114440918, + null + ] + }, + "f2f13ebd94865bf28ae7eeea4ef29cdf": { + "initial_parameters": { + "gravity_constant": 16.913222, + "plane_inclination": 18.820694, + "mass": 1.214361, + "coulomb_friction_coefficient": 0.2, + "breakaway_friction_coefficient": 6, + "initial_velocity": -4.057459, + "Amplitude": -33.145587, + "Period": 2.499777 + }, + "intervention_time": 3.95, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.057143, + "first_diff": [ + 4.059999942779541, + 3.9800000190734863, + null + ] + }, + "2dec7910f3a25f448d279e1bcd61c763": { + "initial_parameters": { + "gravity_constant": 9.91026, + "plane_inclination": 26.225636, + "mass": 2.905654, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 4, + "initial_velocity": 5.328369, + "Amplitude": -335.719855, + "Period": 0.864871 + }, + "intervention_time": 2.7, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "2fc73742c4235516a326bc749d661972": { + "initial_parameters": { + "gravity_constant": 9.91026, + "plane_inclination": 26.225636, + "mass": 2.905654, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 4, + "initial_velocity": 5.328369, + "Amplitude": -335.719855, + "Period": 0.864871 + }, + "intervention_time": 3.13, + "changed_parameter": "plane_inclination", + "new_value": 46.225636, + "first_diff": [ + 3.180000066757202, + 3.1600000858306885, + 3.1600000858306885 + ] + }, + "b7d2dbe41aef52c58ccf2a7fd13903a5": { + "initial_parameters": { + "gravity_constant": 9.91026, + "plane_inclination": 26.225636, + "mass": 2.905654, + "coulomb_friction_coefficient": 1, + "breakaway_friction_coefficient": 4, + "initial_velocity": 5.328369, + "Amplitude": -335.719855, + "Period": 0.864871 + }, + "intervention_time": 4.08, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.285714, + "first_diff": [ + 4.119999885559082, + 4.099999904632568, + null + ] + }, + "e803fdb9e638552d9050b896808b9b46": { + "initial_parameters": { + "gravity_constant": 10.532979, + "plane_inclination": 17.194497, + "mass": 4.739284, + "coulomb_friction_coefficient": 0.77, + "breakaway_friction_coefficient": 1.095, + "initial_velocity": -1.79651, + "Amplitude": -261.38066, + "Period": 2.818427 + }, + "intervention_time": 2.83, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "c6cc5bbcb54c5d37b6f62836386f8f14": { + "initial_parameters": { + "gravity_constant": 10.532979, + "plane_inclination": 17.194497, + "mass": 4.739284, + "coulomb_friction_coefficient": 0.77, + "breakaway_friction_coefficient": 1.095, + "initial_velocity": -1.79651, + "Amplitude": -261.38066, + "Period": 2.818427 + }, + "intervention_time": 2.92, + "changed_parameter": "gravity_constant", + "new_value": 4.388741, + "first_diff": [ + 2.9600000381469727, + 2.940000057220459, + 2.940000057220459 + ] + }, + "660fdef92d435e0a88bb41b9e958b990": { + "initial_parameters": { + "gravity_constant": 14.760357, + "plane_inclination": 20.525364, + "mass": 4.215503, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -2.637982, + "Amplitude": 213.047532, + "Period": 2.760845 + }, + "intervention_time": 2.96, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "8aa929de582d5cbf83aa53fe9ab0a174": { + "initial_parameters": { + "gravity_constant": 14.760357, + "plane_inclination": 20.525364, + "mass": 4.215503, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -2.637982, + "Amplitude": 213.047532, + "Period": 2.760845 + }, + "intervention_time": 3.39, + "changed_parameter": "plane_inclination", + "new_value": 40.525364, + "first_diff": [ + null, + 3.4200000762939453, + 3.4200000762939453 + ] + }, + "3273b1d94d045773a3ed5d180a4511dc": { + "initial_parameters": { + "gravity_constant": 14.760357, + "plane_inclination": 20.525364, + "mass": 4.215503, + "coulomb_friction_coefficient": 3, + "breakaway_friction_coefficient": 6, + "initial_velocity": -2.637982, + "Amplitude": 213.047532, + "Period": 2.760845 + }, + "intervention_time": 4.77, + "changed_parameter": "breakaway_friction_coefficient", + "new_value": 3, + "first_diff": [ + 5.539999961853027, + 5.539999961853027, + null + ] + }, + "9646ae5511015044b647887e96eb8dca": { + "initial_parameters": { + "gravity_constant": 17.048035, + "plane_inclination": 44.220498, + "mass": 4.286929, + "coulomb_friction_coefficient": 0.5, + "breakaway_friction_coefficient": 6, + "initial_velocity": -4.512019, + "Amplitude": -352.943723, + "Period": 1.378285 + }, + "intervention_time": 3.09, + "changed_parameter": "gravity_constant", + "new_value": 28, + "first_diff": [ + 3.240000009536743, + 3.119999885559082, + 3.119999885559082 + ] + }, + "e4156c6755bc5b86ab4fb7c158deb515": { + "initial_parameters": { + "gravity_constant": 17.048035, + "plane_inclination": 44.220498, + "mass": 4.286929, + "coulomb_friction_coefficient": 0.5, + "breakaway_friction_coefficient": 6, + "initial_velocity": -4.512019, + "Amplitude": -352.943723, + "Period": 1.378285 + }, + "intervention_time": 4.38, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 1.25, + "first_diff": [ + 4.460000038146973, + 4.400000095367432, + null + ] + }, + "e40e7855b79058599803ffe83e69df2f": { + "initial_parameters": { + "gravity_constant": 17.048035, + "plane_inclination": 44.220498, + "mass": 4.286929, + "coulomb_friction_coefficient": 0.5, + "breakaway_friction_coefficient": 6, + "initial_velocity": -4.512019, + "Amplitude": -352.943723, + "Period": 1.378285 + }, + "intervention_time": 4.47, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.142857, + "first_diff": [ + 4.639999866485596, + 4.5, + null + ] + }, + "f3198e91ff7d593b8b1d00e99b154708": { + "initial_parameters": { + "gravity_constant": 7.050139, + "plane_inclination": 25.19604, + "mass": 4.133597, + "coulomb_friction_coefficient": 0.25, + "breakaway_friction_coefficient": 3.5, + "initial_velocity": 8.971472, + "Amplitude": -323.794249, + "Period": 0.812169 + }, + "intervention_time": 3.22, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "d2354df65fff5ddf9e1b8e5d138282bc": { + "initial_parameters": { + "gravity_constant": 7.050139, + "plane_inclination": 25.19604, + "mass": 4.133597, + "coulomb_friction_coefficient": 0.25, + "breakaway_friction_coefficient": 3.5, + "initial_velocity": 8.971472, + "Amplitude": -323.794249, + "Period": 0.812169 + }, + "intervention_time": 3.22, + "changed_parameter": "gravity_constant", + "new_value": 11.632729, + "first_diff": [ + 3.5399999618530273, + 3.240000009536743, + 3.240000009536743 + ] + }, + "bb5c824b74c65fd0bfdeeb1b5969e46c": { + "initial_parameters": { + "gravity_constant": 7.050139, + "plane_inclination": 25.19604, + "mass": 4.133597, + "coulomb_friction_coefficient": 0.25, + "breakaway_friction_coefficient": 3.5, + "initial_velocity": 8.971472, + "Amplitude": -323.794249, + "Period": 0.812169 + }, + "intervention_time": 3.65, + "changed_parameter": "plane_inclination", + "new_value": 45.19604, + "first_diff": [ + 3.7799999713897705, + 3.680000066757202, + 3.680000066757202 + ] + }, + "e2cc7f4083515d948057472541e749f9": { + "initial_parameters": { + "gravity_constant": 7.050139, + "plane_inclination": 25.19604, + "mass": 4.133597, + "coulomb_friction_coefficient": 0.25, + "breakaway_friction_coefficient": 3.5, + "initial_velocity": 8.971472, + "Amplitude": -323.794249, + "Period": 0.812169 + }, + "intervention_time": 4.51, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.7, + "first_diff": [ + 4.639999866485596, + 4.539999961853027, + null + ] + }, + "ead1bd72fabc53758d90a08e1dcd2507": { + "initial_parameters": { + "gravity_constant": 7.050139, + "plane_inclination": 25.19604, + "mass": 4.133597, + "coulomb_friction_coefficient": 0.25, + "breakaway_friction_coefficient": 3.5, + "initial_velocity": 8.971472, + "Amplitude": -323.794249, + "Period": 0.812169 + }, + "intervention_time": 4.6, + "changed_parameter": "coulomb_friction_coefficient", + "new_value": 0.071429, + "first_diff": [ + 4.980000019073486, + 4.619999885559082, + null + ] + }, + "d8108f5dafb05591a0165983da77f78c": { + "initial_parameters": { + "gravity_constant": 18.431591, + "plane_inclination": 30.263206, + "mass": 6.684347, + "coulomb_friction_coefficient": 1.8, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.335309, + "Amplitude": 258.220617, + "Period": 2.619658 + }, + "intervention_time": 3.35, + "changed_parameter": "no_parameter_change", + "new_value": null, + "first_diff": [ + null, + null, + null + ] + }, + "e459f3032a4152a496af1745a8b75c10": { + "initial_parameters": { + "gravity_constant": 18.431591, + "plane_inclination": 30.263206, + "mass": 6.684347, + "coulomb_friction_coefficient": 1.8, + "breakaway_friction_coefficient": 6, + "initial_velocity": -3.335309, + "Amplitude": 258.220617, + "Period": 2.619658 + }, + "intervention_time": 3.87, + "changed_parameter": "plane_inclination", + "new_value": 10.263206, + "first_diff": [ + null, + 3.9000000953674316, + 3.9000000953674316 + ] + } + } + }, + "environment_name": "InclinedPlane" +} diff --git a/questions/MassSlide/sample_manifest.json b/questions/MassSlide/sample_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..ba4c25f20a2dcb4b571a754f66eea8121cc888e6 --- /dev/null +++ b/questions/MassSlide/sample_manifest.json @@ -0,0 +1,2833 @@ +{ + "frost_01234": [ + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "31dcf19a3b73436f9e844e7822ac52c3", + "23cca9a4b30252e89b56e757b3b6b0f0", + "914b438477dc4740af3fa7291646e9ce", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "c56def5d135e40c290812174a4d3a066", + "52cb93e7e69d4abf85d6db293440f76c", + "472873131567464dab68c8060f142ec0", + "96814fa562ac4bac8cdd1ace014be742", + "cc258dc235904463b1dba49ca7c1b97f", + "e803fdb9e638552d9050b896808b9b46" + ], + "test_samples": [ + "1f0a940c521643948cd6a4a3f8c15067", + "23cca9a4b30252e89b56e757b3b6b0f0", + "450660a3c2d24f5bae3eeec80638199c", + "509ca665d1854fa596c901a5468197ab", + "95357a192b404d9d942b699a96f3f91d", + "a3dc646b3f7647079bfb49c9acaa18e6", + "b1fb837092144b5cae1c2a952a2ea10c", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "e803fdb9e638552d9050b896808b9b46" + ], + "test_samples_hashes": [ + "fcb9df154efc9f87f1877b165535cc23", + "29523f34cd781b060565e7e04af96b5c", + "0278fcdfaf128057578d409d99cc85de", + "edf4a264ac6a522c53e7ebd3aec2e085", + "9f0e10f22186a49d6e156a0074913a6c", + "abdfb62459ed77149a2e3dff9e494c72", + "e6161e86609d22c3e2b4a980d1aee886", + "6f47662764e804f36d5be292e31099eb", + "eacc4946d6f5b022256b47ba95bfc884", + "258ce11aaf148879738d75ac5c66e08a" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "6799c692b3dceef16296cfb8a7178d8d98cc7e63e91f9140c5fe1112397da8ed", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.4, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "31dcf19a3b73436f9e844e7822ac52c3", + "2dec7910f3a25f448d279e1bcd61c763", + "852557979fcf490298f6359bfb41fbd3", + "38482b8747bc4bfe88ac617356e9c6fe", + "253c13b3eacb4fd2a7778e99a728fd0f", + "9f2a4f73dc22406a8e8277f1a83ed265", + "0dc61f09ac3a57cba3e986abd378a58c", + "22eb60ad8f495387903e93a80cf12294", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "52cb93e7e69d4abf85d6db293440f76c" + ], + "test_samples": [ + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2f00e9a90e3c4484aa48107c021745aa", + "38482b8747bc4bfe88ac617356e9c6fe", + "4b91015695f94331ba68ab2f9bc8feb1", + "633f8b94b4a64afab5c4183b67987092", + "d950be4ceaf855e4b99175c9b68e37f5", + "e4156c6755bc5b86ab4fb7c158deb515", + "e6e67852460340c291d3b2d33978e955", + "f664534ad3bb423eae4c699ad1d1ad79" + ], + "test_samples_hashes": [ + "c8c2849f9ced3d67e82fa39f6b71ffbd", + "9f7d4de458d1c5a1a9e72004056e2663", + "f6a0a75c86fcd1fe59005083af35e754", + "1bf32ba6559ba683e01da61006480882", + "c4cdd6d4e155b756f2c56d0daf860e27", + "d3bacc516c99297eb00354bca60ba401", + "1d0e6402134fc1093ad58701a2c95083", + "4cb15dd030ef6fa6dee844e600116d8d", + "988c36d05934128c74294961821599ef", + "7ec41326259c36505116a76220602663" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "b4c5c674cb9dfb31a2dc34906c68c1aa66093a5a4985ecfc79000ddc391be981", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "852557979fcf490298f6359bfb41fbd3", + "3dee5963901d530e9f2fde3c84a55982", + "38482b8747bc4bfe88ac617356e9c6fe", + "bdf601681b8541659ad53f2b52d11526", + "96814fa562ac4bac8cdd1ace014be742", + "a8e9bd401ebe4e9fb6d22cb8c5acdb47", + "f3198e91ff7d593b8b1d00e99b154708", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "d7efe6d51f6c5e6d8edd9ddff98f01f4" + ], + "test_samples": [ + "03cf9a5fd2af43aa8b2413596e24ffa4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "53d61728255b43158f043258bf085cb9", + "97c922e470a6420aaa0db38a7b47ab14", + "9b572270b12942a390d67ff9595a2cbd", + "b9d5d04e374c47d58d820fca9d364a11", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "ffcb3e01741a516091f8da7aeb689274" + ], + "test_samples_hashes": [ + "911393fb9b44bf1e0ae603e8277066d8", + "d0a0934b84401abf6d427c5051cdb5a9", + "87ffc1178bf3675e97c2b9c7268f7ae0", + "cde69e48968965e5ffac6a34a8ae3532", + "5eec7d64ce95c19cce60da50fbac57b1", + "98b81617fc2dfa7a56b351b51a3c4c22", + "1041e14f858ba45a53f9382d423a1bff", + "48f61cda56b12176da48acee13424ecb", + "385969c2db1796fe0788c245f81ea1ea", + "ec12dafabe177c85eb913aeb72956169" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "1795fff5c2b9368a7cb10f7502ee104b07ad2d607f06ece68da8613662af5aa6", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.2, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "afec7fdef34d4256a0060406c9b1cd7c", + "660fdef92d435e0a88bb41b9e958b990", + "7d1520b2be3c531ab4acbe645c9e18cd", + "b6e5860643684682b60e6c146121b2be", + "472873131567464dab68c8060f142ec0", + "c1e94ce33230421a847d36495497409d", + "22eb60ad8f495387903e93a80cf12294", + "d8108f5dafb05591a0165983da77f78c", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "31dcf19a3b73436f9e844e7822ac52c3" + ], + "test_samples": [ + "260750f9d27a4578ba899fcfb02ac653", + "3273b1d94d045773a3ed5d180a4511dc", + "7d1520b2be3c531ab4acbe645c9e18cd", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "bd222ab0a4034903923ada8b7a09816f", + "c1e94ce33230421a847d36495497409d", + "e40e7855b79058599803ffe83e69df2f", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "fb2367ca63f6468bbb95f4ddbd0d99c9" + ], + "test_samples_hashes": [ + "47f716a3f81a37b815ac8edd619a0ceb", + "cdd906c5b95a94a735bd54c97bd58a2a", + "4cfa3a4e5d448cf4fa577012c45d555f", + "ea935ea072de6c0dd32b104da965a06b", + "423f2fec735b03ca2a4f38628bade3a8", + "700d0f17f4c708766823cd85a5c5a833", + "c3c397052ce82a4ff250f2f81f6d7e25", + "570396e18c3280218b96bfaf41280dc0", + "f4928cc375bb9b021a302e566b6c3062", + "f0150020d3dbb1c4558dd1b87ef56135" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "2c8709176b4f21a09212fc4778e6c95711fbef8ac49b595aa42b472938f02d36", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bdf601681b8541659ad53f2b52d11526", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [], + "train_samples": [], + "train_samples_hashes": [], + "test_samples_baselines": [ + "2dec7910f3a25f448d279e1bcd61c763", + "322aa53bc0ae43c08d03196c3fa01dcf", + "31dcf19a3b73436f9e844e7822ac52c3", + "b794f7c284485e548f68445b0c123a3a", + "22eb60ad8f495387903e93a80cf12294", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "4f105c22a575422faa161016192329d2", + "f3198e91ff7d593b8b1d00e99b154708", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "38482b8747bc4bfe88ac617356e9c6fe" + ], + "test_samples": [ + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3ad87b80437b43fda95e551a2efc7b87", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "9646ae5511015044b647887e96eb8dca", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "e2cc7f4083515d948057472541e749f9", + "e8e9190868644fc38aa1113756e04fee", + "ff3a19d5eaad46c1aed4ca4373c26de3" + ], + "test_samples_hashes": [ + "717393ab4afb9d948c8faa1abeb944d5", + "c2e3dade987a5c80aab3c25a1f5f0b21", + "6daa1de025af0c11ca69b85d58be04dc", + "1e8c2a2e489efb1d77918c94cd44b248", + "85c972828adf0db2de946c8143aa3955", + "7dfdef65098f35fb6f5804a59777d9cf", + "3087b5c9008db5fa7b442d4d329c21dc", + "7cd0b94e84b5ae80e347e423efa65420", + "5a5f4d277150617f1f739a6e55cc29e1", + "f27adfcf89db0298da43ee6f363db29e" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": null, + "number_train_samples_per_class": 0, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.0 + }, + "train_test_sample_hash": "22182ae00c573129e762df0c2c49c7379c5d9988f9ad56efde880e96238d6b0e", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e40e7855b79058599803ffe83e69df2f", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.3 + } + } + ], + "fern_01234": [ + { + "train_samples_baselines": [ + "9f2a4f73dc22406a8e8277f1a83ed265", + "660fdef92d435e0a88bb41b9e958b990", + "2713905d0dfa4d56aac01b96531ab3e7", + "b794f7c284485e548f68445b0c123a3a", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f" + ], + "train_samples": [ + "38e9b9cde3d547a9924b3af04070203a", + "660fdef92d435e0a88bb41b9e958b990", + "7b689cd67c2d4021930e364b1c2ed8ca", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "d1141599d62d4cd8a57708873e904583" + ], + "train_samples_hashes": [ + "eea58208b52bebf71aa8d5e56c4b0b07", + "bdad9f29b10b1d4f71121b26b315c1ab", + "aa8b6366a6896dd0bba26cb029df2e33", + "1e8c2a2e489efb1d77918c94cd44b248", + "1f27aef26a69058a1f42a08e58f3dda6" + ], + "test_samples_baselines": [ + "31dcf19a3b73436f9e844e7822ac52c3", + "23cca9a4b30252e89b56e757b3b6b0f0", + "914b438477dc4740af3fa7291646e9ce", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "c56def5d135e40c290812174a4d3a066", + "52cb93e7e69d4abf85d6db293440f76c", + "472873131567464dab68c8060f142ec0", + "96814fa562ac4bac8cdd1ace014be742", + "cc258dc235904463b1dba49ca7c1b97f", + "e803fdb9e638552d9050b896808b9b46" + ], + "test_samples": [ + "1f0a940c521643948cd6a4a3f8c15067", + "23cca9a4b30252e89b56e757b3b6b0f0", + "450660a3c2d24f5bae3eeec80638199c", + "509ca665d1854fa596c901a5468197ab", + "95357a192b404d9d942b699a96f3f91d", + "a3dc646b3f7647079bfb49c9acaa18e6", + "b1fb837092144b5cae1c2a952a2ea10c", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "e803fdb9e638552d9050b896808b9b46" + ], + "test_samples_hashes": [ + "fcb9df154efc9f87f1877b165535cc23", + "29523f34cd781b060565e7e04af96b5c", + "0278fcdfaf128057578d409d99cc85de", + "edf4a264ac6a522c53e7ebd3aec2e085", + "9f0e10f22186a49d6e156a0074913a6c", + "abdfb62459ed77149a2e3dff9e494c72", + "e6161e86609d22c3e2b4a980d1aee886", + "6f47662764e804f36d5be292e31099eb", + "eacc4946d6f5b022256b47ba95bfc884", + "258ce11aaf148879738d75ac5c66e08a" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.0, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "11f7f8622d2a2d9d6464cf16959927fcaecd184d5e7b05fd7d645d5781f19d36", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "6ff497703dc54edb8534f212c5b7c567", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "924138e0d0f9479f8dccc89cc12e589c", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.4, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [ + "d7efe6d51f6c5e6d8edd9ddff98f01f4", + "96814fa562ac4bac8cdd1ace014be742", + "bdf601681b8541659ad53f2b52d11526", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "b6e5860643684682b60e6c146121b2be" + ], + "train_samples": [ + "52d91d5389a652baa478069ea293ca73", + "96814fa562ac4bac8cdd1ace014be742", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "b6a00423859b48e8b634aaf7a4b4208a", + "bd1db72c3cdb481caa778ac3639cccb3" + ], + "train_samples_hashes": [ + "695ec75f3338a7ed84dbeffd50c2298d", + "093451c4ffe836f6991a0837aac5b941", + "6e06d292fc5f0e83e260d6805c3a3667", + "0775335a3ecb8e62c3eb76f6101a1815", + "2830941891adeffb526c72a63271f944" + ], + "test_samples_baselines": [ + "31dcf19a3b73436f9e844e7822ac52c3", + "2dec7910f3a25f448d279e1bcd61c763", + "852557979fcf490298f6359bfb41fbd3", + "38482b8747bc4bfe88ac617356e9c6fe", + "253c13b3eacb4fd2a7778e99a728fd0f", + "9f2a4f73dc22406a8e8277f1a83ed265", + "0dc61f09ac3a57cba3e986abd378a58c", + "22eb60ad8f495387903e93a80cf12294", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "52cb93e7e69d4abf85d6db293440f76c" + ], + "test_samples": [ + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2f00e9a90e3c4484aa48107c021745aa", + "38482b8747bc4bfe88ac617356e9c6fe", + "4b91015695f94331ba68ab2f9bc8feb1", + "633f8b94b4a64afab5c4183b67987092", + "d950be4ceaf855e4b99175c9b68e37f5", + "e4156c6755bc5b86ab4fb7c158deb515", + "e6e67852460340c291d3b2d33978e955", + "f664534ad3bb423eae4c699ad1d1ad79" + ], + "test_samples_hashes": [ + "c8c2849f9ced3d67e82fa39f6b71ffbd", + "9f7d4de458d1c5a1a9e72004056e2663", + "f6a0a75c86fcd1fe59005083af35e754", + "1bf32ba6559ba683e01da61006480882", + "c4cdd6d4e155b756f2c56d0daf860e27", + "d3bacc516c99297eb00354bca60ba401", + "1d0e6402134fc1093ad58701a2c95083", + "4cb15dd030ef6fa6dee844e600116d8d", + "988c36d05934128c74294961821599ef", + "7ec41326259c36505116a76220602663" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "0e177bc9873803361541361838e7944f83cc46fb206b3c526848a747074237fc", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [ + "2713905d0dfa4d56aac01b96531ab3e7", + "b4591a30e59f5c4897a79165f3d336ac", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "1e4d0d6cd30859d491b34dc0200b99a7" + ], + "train_samples": [ + "223c934b74a24e5f9df19997035eef2f", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "e0934b52a37759e3a90cd51c68f250e5" + ], + "train_samples_hashes": [ + "006d163d2ab583e31a904123fdd046d0", + "a2b17a57de69fac673dac3179ab5c83d", + "cf99288aa0ffe147b58431d2f616b3a9", + "cffa3fdf64e42d8be45cfebcbd63ee46", + "5cb01e446967ea0520dfbbf1e92eb801" + ], + "test_samples_baselines": [ + "852557979fcf490298f6359bfb41fbd3", + "3dee5963901d530e9f2fde3c84a55982", + "38482b8747bc4bfe88ac617356e9c6fe", + "bdf601681b8541659ad53f2b52d11526", + "96814fa562ac4bac8cdd1ace014be742", + "a8e9bd401ebe4e9fb6d22cb8c5acdb47", + "f3198e91ff7d593b8b1d00e99b154708", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "d7efe6d51f6c5e6d8edd9ddff98f01f4" + ], + "test_samples": [ + "03cf9a5fd2af43aa8b2413596e24ffa4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "53d61728255b43158f043258bf085cb9", + "97c922e470a6420aaa0db38a7b47ab14", + "9b572270b12942a390d67ff9595a2cbd", + "b9d5d04e374c47d58d820fca9d364a11", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "ffcb3e01741a516091f8da7aeb689274" + ], + "test_samples_hashes": [ + "911393fb9b44bf1e0ae603e8277066d8", + "d0a0934b84401abf6d427c5051cdb5a9", + "87ffc1178bf3675e97c2b9c7268f7ae0", + "cde69e48968965e5ffac6a34a8ae3532", + "5eec7d64ce95c19cce60da50fbac57b1", + "98b81617fc2dfa7a56b351b51a3c4c22", + "1041e14f858ba45a53f9382d423a1bff", + "48f61cda56b12176da48acee13424ecb", + "385969c2db1796fe0788c245f81ea1ea", + "ec12dafabe177c85eb913aeb72956169" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "fe08147ad280fcddb2271c2f949188cbcf2cf55805f9c0ad110f3b014d004767", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.2, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "115036fcb4cb40278192aeaec3c948ba", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "2dec7910f3a25f448d279e1bcd61c763", + "924138e0d0f9479f8dccc89cc12e589c", + "f3198e91ff7d593b8b1d00e99b154708" + ], + "train_samples": [ + "115036fcb4cb40278192aeaec3c948ba", + "208ef509e8b04166ae5333a2877f1295", + "2fc73742c4235516a326bc749d661972", + "9e10b79464c24f36a53f252be6541d5f", + "d2354df65fff5ddf9e1b8e5d138282bc" + ], + "train_samples_hashes": [ + "a3d845d0ffd80def20e5129bfe8a7ea2", + "a44a8f5df70186b459b4a2564f89e367", + "717393ab4afb9d948c8faa1abeb944d5", + "3a6ea1d1c0a69068dcc5c61e6f645066", + "0d21d8f3c8d4e08b0762b8951b3374b3" + ], + "test_samples_baselines": [ + "afec7fdef34d4256a0060406c9b1cd7c", + "660fdef92d435e0a88bb41b9e958b990", + "7d1520b2be3c531ab4acbe645c9e18cd", + "b6e5860643684682b60e6c146121b2be", + "472873131567464dab68c8060f142ec0", + "c1e94ce33230421a847d36495497409d", + "22eb60ad8f495387903e93a80cf12294", + "d8108f5dafb05591a0165983da77f78c", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "31dcf19a3b73436f9e844e7822ac52c3" + ], + "test_samples": [ + "260750f9d27a4578ba899fcfb02ac653", + "3273b1d94d045773a3ed5d180a4511dc", + "7d1520b2be3c531ab4acbe645c9e18cd", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "bd222ab0a4034903923ada8b7a09816f", + "c1e94ce33230421a847d36495497409d", + "e40e7855b79058599803ffe83e69df2f", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "fb2367ca63f6468bbb95f4ddbd0d99c9" + ], + "test_samples_hashes": [ + "47f716a3f81a37b815ac8edd619a0ceb", + "cdd906c5b95a94a735bd54c97bd58a2a", + "4cfa3a4e5d448cf4fa577012c45d555f", + "ea935ea072de6c0dd32b104da965a06b", + "423f2fec735b03ca2a4f38628bade3a8", + "700d0f17f4c708766823cd85a5c5a833", + "c3c397052ce82a4ff250f2f81f6d7e25", + "570396e18c3280218b96bfaf41280dc0", + "f4928cc375bb9b021a302e566b6c3062", + "f0150020d3dbb1c4558dd1b87ef56135" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "7a097682902340a904b33e388b7474b9e26ddf3f721c64c1d908e29d4104522c", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "322aa53bc0ae43c08d03196c3fa01dcf", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bdf601681b8541659ad53f2b52d11526", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "7d1520b2be3c531ab4acbe645c9e18cd", + "7d1520b2be3c531ab4acbe645c9e18cd", + "a8e9bd401ebe4e9fb6d22cb8c5acdb47", + "c1e94ce33230421a847d36495497409d", + "0dc61f09ac3a57cba3e986abd378a58c" + ], + "train_samples": [ + "0d79edc07243565d8d8b5380ef81e340", + "402307421702585c86a2ff40516deca2", + "b9d5d04e374c47d58d820fca9d364a11", + "c1e94ce33230421a847d36495497409d", + "f2f13ebd94865bf28ae7eeea4ef29cdf" + ], + "train_samples_hashes": [ + "80bc81c5eb48eb5d2ec18c66f060fe46", + "8a106e0f4c530f51c12d5dd98af10671", + "98b81617fc2dfa7a56b351b51a3c4c22", + "700d0f17f4c708766823cd85a5c5a833", + "4a58e5b8d56eafc0f928c6e16c643003" + ], + "test_samples_baselines": [ + "2dec7910f3a25f448d279e1bcd61c763", + "322aa53bc0ae43c08d03196c3fa01dcf", + "31dcf19a3b73436f9e844e7822ac52c3", + "b794f7c284485e548f68445b0c123a3a", + "22eb60ad8f495387903e93a80cf12294", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "4f105c22a575422faa161016192329d2", + "f3198e91ff7d593b8b1d00e99b154708", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "38482b8747bc4bfe88ac617356e9c6fe" + ], + "test_samples": [ + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3ad87b80437b43fda95e551a2efc7b87", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "9646ae5511015044b647887e96eb8dca", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "e2cc7f4083515d948057472541e749f9", + "e8e9190868644fc38aa1113756e04fee", + "ff3a19d5eaad46c1aed4ca4373c26de3" + ], + "test_samples_hashes": [ + "717393ab4afb9d948c8faa1abeb944d5", + "c2e3dade987a5c80aab3c25a1f5f0b21", + "6daa1de025af0c11ca69b85d58be04dc", + "1e8c2a2e489efb1d77918c94cd44b248", + "85c972828adf0db2de946c8143aa3955", + "7dfdef65098f35fb6f5804a59777d9cf", + "3087b5c9008db5fa7b442d4d329c21dc", + "7cd0b94e84b5ae80e347e423efa65420", + "5a5f4d277150617f1f739a6e55cc29e1", + "f27adfcf89db0298da43ee6f363db29e" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 1, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "5eed0b0ab2b3e208cd44b2683b95802b7b1d25e67bedc993fd69e43efe777bc3", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e40e7855b79058599803ffe83e69df2f", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "ead1bd72fabc53758d90a08e1dcd2507", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.3 + } + } + ], + "gentle_01234": [ + { + "train_samples_baselines": [ + "852557979fcf490298f6359bfb41fbd3", + "0dc61f09ac3a57cba3e986abd378a58c", + "2dec7910f3a25f448d279e1bcd61c763", + "4f105c22a575422faa161016192329d2", + "3dee5963901d530e9f2fde3c84a55982", + "b4591a30e59f5c4897a79165f3d336ac", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "bdf601681b8541659ad53f2b52d11526", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "2713905d0dfa4d56aac01b96531ab3e7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "bdf601681b8541659ad53f2b52d11526", + "22eb60ad8f495387903e93a80cf12294", + "f3198e91ff7d593b8b1d00e99b154708" + ], + "train_samples": [ + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "2fc73742c4235516a326bc749d661972", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dee5963901d530e9f2fde3c84a55982", + "4936a5b652995d1cbb5987f3c7ee6866", + "833168b23c474c3f95c1ba4d9fff6176", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "97f285a3809940f68129ed3a9d1d9b26", + "9e2271b4625541f7964a03844c658511", + "a788bcbddb3e4143afc049ae6fd613ed", + "b6a00423859b48e8b634aaf7a4b4208a", + "bdf601681b8541659ad53f2b52d11526", + "e4156c6755bc5b86ab4fb7c158deb515", + "f3198e91ff7d593b8b1d00e99b154708" + ], + "train_samples_hashes": [ + "911393fb9b44bf1e0ae603e8277066d8", + "3aad474e508a4332f1cbea5e592f08ea", + "717393ab4afb9d948c8faa1abeb944d5", + "be56fc0b8e91df425c208d2fc13d7349", + "26468440501aa53665093719ddd3f007", + "8bbd58b7e57baf7a421b38324475239b", + "425f8e0929b7fdddec4c7aa6982338db", + "1e8c2a2e489efb1d77918c94cd44b248", + "a30b5b00b5368b79cb13dbf2eae47b07", + "b8827508f18d571c18ad4345357f8c5f", + "a4cbaac4e7dd1e19fe2bb5c41a814c2c", + "0775335a3ecb8e62c3eb76f6101a1815", + "02fccb048ef7fb32d53361dc56ac8748", + "4cb15dd030ef6fa6dee844e600116d8d", + "318e2ab6ec33d6dc88af05059f1c205b" + ], + "test_samples_baselines": [ + "31dcf19a3b73436f9e844e7822ac52c3", + "23cca9a4b30252e89b56e757b3b6b0f0", + "914b438477dc4740af3fa7291646e9ce", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "c56def5d135e40c290812174a4d3a066", + "52cb93e7e69d4abf85d6db293440f76c", + "472873131567464dab68c8060f142ec0", + "96814fa562ac4bac8cdd1ace014be742", + "cc258dc235904463b1dba49ca7c1b97f", + "e803fdb9e638552d9050b896808b9b46" + ], + "test_samples": [ + "1f0a940c521643948cd6a4a3f8c15067", + "23cca9a4b30252e89b56e757b3b6b0f0", + "450660a3c2d24f5bae3eeec80638199c", + "509ca665d1854fa596c901a5468197ab", + "95357a192b404d9d942b699a96f3f91d", + "a3dc646b3f7647079bfb49c9acaa18e6", + "b1fb837092144b5cae1c2a952a2ea10c", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "e803fdb9e638552d9050b896808b9b46" + ], + "test_samples_hashes": [ + "fcb9df154efc9f87f1877b165535cc23", + "29523f34cd781b060565e7e04af96b5c", + "0278fcdfaf128057578d409d99cc85de", + "edf4a264ac6a522c53e7ebd3aec2e085", + "9f0e10f22186a49d6e156a0074913a6c", + "abdfb62459ed77149a2e3dff9e494c72", + "e6161e86609d22c3e2b4a980d1aee886", + "6f47662764e804f36d5be292e31099eb", + "eacc4946d6f5b022256b47ba95bfc884", + "258ce11aaf148879738d75ac5c66e08a" + ], + "seed": 0, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "983a576ebeccac02ba07a78e54f26e7c4e2e22bbd9826b0e9bc930fdd875eac0", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "477fa4f8ca014f559465bdc2c2318c78", + "4b91015695f94331ba68ab2f9bc8feb1", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "924138e0d0f9479f8dccc89cc12e589c", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.4, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [ + "b794f7c284485e548f68445b0c123a3a", + "472873131567464dab68c8060f142ec0", + "914b438477dc4740af3fa7291646e9ce", + "a8e9bd401ebe4e9fb6d22cb8c5acdb47", + "d7efe6d51f6c5e6d8edd9ddff98f01f4", + "914b438477dc4740af3fa7291646e9ce", + "660fdef92d435e0a88bb41b9e958b990", + "924138e0d0f9479f8dccc89cc12e589c", + "bdf601681b8541659ad53f2b52d11526", + "b6e5860643684682b60e6c146121b2be", + "a8e9bd401ebe4e9fb6d22cb8c5acdb47", + "cc258dc235904463b1dba49ca7c1b97f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "f3198e91ff7d593b8b1d00e99b154708" + ], + "train_samples": [ + "1dbde2fdb75b5ca294baa80857327b67", + "3ba3a08aac1441ab9f7062076ab2e050", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "52d91d5389a652baa478069ea293ca73", + "5cff394d93ba45fc950eb525d975c078", + "8aa929de582d5cbf83aa53fe9ab0a174", + "924138e0d0f9479f8dccc89cc12e589c", + "97c922e470a6420aaa0db38a7b47ab14", + "b6e5860643684682b60e6c146121b2be", + "b9d5d04e374c47d58d820fca9d364a11", + "bd084d19f2444af8a6f98ec7815d2882", + "cc910fa7368c5587804f005866a22079", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "ead1bd72fabc53758d90a08e1dcd2507" + ], + "train_samples_hashes": [ + "8c7ac627e25a9b3e9980414b56764da4", + "e5ba8922218dd2839d29d948257459dd", + "0278fcdfaf128057578d409d99cc85de", + "7b1c46eae667792fb1c46e960dc99993", + "695ec75f3338a7ed84dbeffd50c2298d", + "a558ed4f0e1075ea1aad7f0f57281e41", + "da17d9e56c653e56ec9675171884e429", + "fbdb5171316622c1063d9fba7c5b081c", + "cde69e48968965e5ffac6a34a8ae3532", + "2c5d53ece8f1fa28d30ef0ce13024768", + "98b81617fc2dfa7a56b351b51a3c4c22", + "5aef2efd606ea97c48461efac8cfe20e", + "0e8e00c54151068a7896ec46ef19000e", + "385969c2db1796fe0788c245f81ea1ea", + "98cdbe182bdd85da4d92af494499240c" + ], + "test_samples_baselines": [ + "31dcf19a3b73436f9e844e7822ac52c3", + "2dec7910f3a25f448d279e1bcd61c763", + "852557979fcf490298f6359bfb41fbd3", + "38482b8747bc4bfe88ac617356e9c6fe", + "253c13b3eacb4fd2a7778e99a728fd0f", + "9f2a4f73dc22406a8e8277f1a83ed265", + "0dc61f09ac3a57cba3e986abd378a58c", + "22eb60ad8f495387903e93a80cf12294", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "52cb93e7e69d4abf85d6db293440f76c" + ], + "test_samples": [ + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2f00e9a90e3c4484aa48107c021745aa", + "38482b8747bc4bfe88ac617356e9c6fe", + "4b91015695f94331ba68ab2f9bc8feb1", + "633f8b94b4a64afab5c4183b67987092", + "d950be4ceaf855e4b99175c9b68e37f5", + "e4156c6755bc5b86ab4fb7c158deb515", + "e6e67852460340c291d3b2d33978e955", + "f664534ad3bb423eae4c699ad1d1ad79" + ], + "test_samples_hashes": [ + "c8c2849f9ced3d67e82fa39f6b71ffbd", + "9f7d4de458d1c5a1a9e72004056e2663", + "f6a0a75c86fcd1fe59005083af35e754", + "1bf32ba6559ba683e01da61006480882", + "c4cdd6d4e155b756f2c56d0daf860e27", + "d3bacc516c99297eb00354bca60ba401", + "1d0e6402134fc1093ad58701a2c95083", + "4cb15dd030ef6fa6dee844e600116d8d", + "988c36d05934128c74294961821599ef", + "7ec41326259c36505116a76220602663" + ], + "seed": 1, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "38d835e7db029355c5d68e32511a44f20b54c931336a0c1ebd0b4c0b2af62262", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "337efe3f652b4c12b24b70a63f25adff", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "4936a5b652995d1cbb5987f3c7ee6866", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e40e7855b79058599803ffe83e69df2f", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.2, + "correlation_nn": 0.3 + } + }, + { + "train_samples_baselines": [ + "7d1520b2be3c531ab4acbe645c9e18cd", + "4f105c22a575422faa161016192329d2", + "253c13b3eacb4fd2a7778e99a728fd0f", + "31dcf19a3b73436f9e844e7822ac52c3", + "7d1520b2be3c531ab4acbe645c9e18cd", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "52cb93e7e69d4abf85d6db293440f76c", + "914b438477dc4740af3fa7291646e9ce", + "322aa53bc0ae43c08d03196c3fa01dcf", + "31dcf19a3b73436f9e844e7822ac52c3", + "afec7fdef34d4256a0060406c9b1cd7c", + "c56def5d135e40c290812174a4d3a066", + "1e4d0d6cd30859d491b34dc0200b99a7", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "22eb60ad8f495387903e93a80cf12294" + ], + "train_samples": [ + "0d79edc07243565d8d8b5380ef81e340", + "1d716244585943cc9e58b841706c0a36", + "337efe3f652b4c12b24b70a63f25adff", + "3ad87b80437b43fda95e551a2efc7b87", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "52cb93e7e69d4abf85d6db293440f76c", + "59b2dd9b949c42e7ab404612b867f051", + "825bb775cf154fdb983653ce20d329dc", + "8840791a44ca45f2b1e72ee59093cef4", + "afec7fdef34d4256a0060406c9b1cd7c", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "e40e7855b79058599803ffe83e69df2f" + ], + "train_samples_hashes": [ + "80bc81c5eb48eb5d2ec18c66f060fe46", + "116e6256a44c78a215649d8424c764ba", + "1f5aac387ceaacad16f1ba18eada37a4", + "6daa1de025af0c11ca69b85d58be04dc", + "8a106e0f4c530f51c12d5dd98af10671", + "1d9f83b3539e5a79fade52f3c72e9ab0", + "a9cc86f19309fc05fdef3a4fc2de77d6", + "85e2764df08019ea6685e06982ba4b4f", + "8034c71adb97fc649df2021cca124f95", + "e1221f8c6f6d1ecc6bf568022ba0796d", + "99654a2b38c486d209c90d8a42d4e431", + "4cf0225133f9691c3e863ab18a5f2bb3", + "8fe02ddec1e7c8b13f5eabdd5fa43280", + "cffa3fdf64e42d8be45cfebcbd63ee46", + "c3c397052ce82a4ff250f2f81f6d7e25" + ], + "test_samples_baselines": [ + "852557979fcf490298f6359bfb41fbd3", + "3dee5963901d530e9f2fde3c84a55982", + "38482b8747bc4bfe88ac617356e9c6fe", + "bdf601681b8541659ad53f2b52d11526", + "96814fa562ac4bac8cdd1ace014be742", + "a8e9bd401ebe4e9fb6d22cb8c5acdb47", + "f3198e91ff7d593b8b1d00e99b154708", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "d7efe6d51f6c5e6d8edd9ddff98f01f4" + ], + "test_samples": [ + "03cf9a5fd2af43aa8b2413596e24ffa4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "53d61728255b43158f043258bf085cb9", + "97c922e470a6420aaa0db38a7b47ab14", + "9b572270b12942a390d67ff9595a2cbd", + "b9d5d04e374c47d58d820fca9d364a11", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "ffcb3e01741a516091f8da7aeb689274" + ], + "test_samples_hashes": [ + "911393fb9b44bf1e0ae603e8277066d8", + "d0a0934b84401abf6d427c5051cdb5a9", + "87ffc1178bf3675e97c2b9c7268f7ae0", + "cde69e48968965e5ffac6a34a8ae3532", + "5eec7d64ce95c19cce60da50fbac57b1", + "98b81617fc2dfa7a56b351b51a3c4c22", + "1041e14f858ba45a53f9382d423a1bff", + "48f61cda56b12176da48acee13424ecb", + "385969c2db1796fe0788c245f81ea1ea", + "ec12dafabe177c85eb913aeb72956169" + ], + "seed": 2, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "8d171b58be5ee86560e8c18f33af7894b0c14a92311bc37f6e9e1cba4a5fcd6a", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3273b1d94d045773a3ed5d180a4511dc", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dee5963901d530e9f2fde3c84a55982", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52d91d5389a652baa478069ea293ca73", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "833168b23c474c3f95c1ba4d9fff6176", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97f285a3809940f68129ed3a9d1d9b26", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e0934b52a37759e3a90cd51c68f250e5", + "e2cc7f4083515d948057472541e749f9", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ff3a19d5eaad46c1aed4ca4373c26de3" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.0, + "euclidean_centroid": 0.2, + "correlation_nn": 0.1 + } + }, + { + "train_samples_baselines": [ + "a8e9bd401ebe4e9fb6d22cb8c5acdb47", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "2713905d0dfa4d56aac01b96531ab3e7", + "322aa53bc0ae43c08d03196c3fa01dcf", + "9f2a4f73dc22406a8e8277f1a83ed265", + "9f2a4f73dc22406a8e8277f1a83ed265", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "52cb93e7e69d4abf85d6db293440f76c", + "b794f7c284485e548f68445b0c123a3a", + "2dec7910f3a25f448d279e1bcd61c763", + "c56def5d135e40c290812174a4d3a066", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "f3198e91ff7d593b8b1d00e99b154708" + ], + "train_samples": [ + "00af86611bf0452986e018425db6cdd2", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "2713905d0dfa4d56aac01b96531ab3e7", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "633f8b94b4a64afab5c4183b67987092", + "6ff497703dc54edb8534f212c5b7c567", + "8ffce32ebf95402db33ab77ea1fe5f12", + "a3dc646b3f7647079bfb49c9acaa18e6", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "cfcf113e827240a6b7e7a4be86078864", + "d913f421c0964aea99408cec015c64c1", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e2cc7f4083515d948057472541e749f9" + ], + "train_samples_hashes": [ + "0e4b8232db2bfc1db623f7fe255a9e36", + "b2fd2ce49c308dfdd3b0ab309fa8e33c", + "8f3252478336ac13317e41fe351d861d", + "1c34914631acc2b73a0f8a8f473d4557", + "8eeca15cda7e4f0e8652c3ba0d704f71", + "d3bacc516c99297eb00354bca60ba401", + "b4c9d0e281a2bf124b7908e603c77505", + "5770963896b1f72c958fafb47d6b7233", + "abdfb62459ed77149a2e3dff9e494c72", + "54bc13209359a56d5ca1757ba140ef8f", + "75c5529a24c62c82a061c4a13b78faaa", + "c96b86d4ccfaacdef36a59b38587aae5", + "5e68d80623386912e06fcdb7dd6ea844", + "48f61cda56b12176da48acee13424ecb", + "7cd0b94e84b5ae80e347e423efa65420" + ], + "test_samples_baselines": [ + "afec7fdef34d4256a0060406c9b1cd7c", + "660fdef92d435e0a88bb41b9e958b990", + "7d1520b2be3c531ab4acbe645c9e18cd", + "b6e5860643684682b60e6c146121b2be", + "472873131567464dab68c8060f142ec0", + "c1e94ce33230421a847d36495497409d", + "22eb60ad8f495387903e93a80cf12294", + "d8108f5dafb05591a0165983da77f78c", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "31dcf19a3b73436f9e844e7822ac52c3" + ], + "test_samples": [ + "260750f9d27a4578ba899fcfb02ac653", + "3273b1d94d045773a3ed5d180a4511dc", + "7d1520b2be3c531ab4acbe645c9e18cd", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "bd222ab0a4034903923ada8b7a09816f", + "c1e94ce33230421a847d36495497409d", + "e40e7855b79058599803ffe83e69df2f", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "fb2367ca63f6468bbb95f4ddbd0d99c9" + ], + "test_samples_hashes": [ + "47f716a3f81a37b815ac8edd619a0ceb", + "cdd906c5b95a94a735bd54c97bd58a2a", + "4cfa3a4e5d448cf4fa577012c45d555f", + "ea935ea072de6c0dd32b104da965a06b", + "423f2fec735b03ca2a4f38628bade3a8", + "700d0f17f4c708766823cd85a5c5a833", + "c3c397052ce82a4ff250f2f81f6d7e25", + "570396e18c3280218b96bfaf41280dc0", + "f4928cc375bb9b021a302e566b6c3062", + "f0150020d3dbb1c4558dd1b87ef56135" + ], + "seed": 3, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.1, + "correlation_nn": 0.2 + }, + "train_test_sample_hash": "4763b98b1f6f4b13d2b2952569a6924dd59dd917250eaabbfed7c7287750783d", + "other_samples": [ + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "07fd7c347c7354bbb738d7765a2fea14", + "0d79edc07243565d8d8b5380ef81e340", + "0dc61f09ac3a57cba3e986abd378a58c", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2f00e9a90e3c4484aa48107c021745aa", + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ad87b80437b43fda95e551a2efc7b87", + "3ba3a08aac1441ab9f7062076ab2e050", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5c6ece223a1a46f6a1987449154ff138", + "5cff394d93ba45fc950eb525d975c078", + "660fdef92d435e0a88bb41b9e958b990", + "7b689cd67c2d4021930e364b1c2ed8ca", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "9646ae5511015044b647887e96eb8dca", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9b572270b12942a390d67ff9595a2cbd", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "afec7fdef34d4256a0060406c9b1cd7c", + "b1fb837092144b5cae1c2a952a2ea10c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bdf601681b8541659ad53f2b52d11526", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "cc910fa7368c5587804f005866a22079", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d8108f5dafb05591a0165983da77f78c", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e0934b52a37759e3a90cd51c68f250e5", + "e4156c6755bc5b86ab4fb7c158deb515", + "e430606893ee431087b189d71cd29ac5", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "e8e9190868644fc38aa1113756e04fee", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "f664534ad3bb423eae4c699ad1d1ad79", + "ff3a19d5eaad46c1aed4ca4373c26de3", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.2 + } + }, + { + "train_samples_baselines": [ + "1e4d0d6cd30859d491b34dc0200b99a7", + "0dc61f09ac3a57cba3e986abd378a58c", + "660fdef92d435e0a88bb41b9e958b990", + "924138e0d0f9479f8dccc89cc12e589c", + "1e4d0d6cd30859d491b34dc0200b99a7", + "cac45b19acaf43b190ce566ea5b23d9a", + "b6e5860643684682b60e6c146121b2be", + "96814fa562ac4bac8cdd1ace014be742", + "472873131567464dab68c8060f142ec0", + "e803fdb9e638552d9050b896808b9b46", + "c1e94ce33230421a847d36495497409d", + "d8108f5dafb05591a0165983da77f78c", + "dd028d3ab7fb4aae84694d21ecf4c201", + "96814fa562ac4bac8cdd1ace014be742", + "52cb93e7e69d4abf85d6db293440f76c" + ], + "train_samples": [ + "07fd7c347c7354bbb738d7765a2fea14", + "0dc61f09ac3a57cba3e986abd378a58c", + "3273b1d94d045773a3ed5d180a4511dc", + "3c9308c66d644b07be5b2c3ee6ffa4e6", + "3cdf2c46f20e55c0b081c3745fbf8c40", + "5c6ece223a1a46f6a1987449154ff138", + "9829e3384fae4a77ae3bc21c6c8a6aea", + "9b572270b12942a390d67ff9595a2cbd", + "b1fb837092144b5cae1c2a952a2ea10c", + "c6cc5bbcb54c5d37b6f62836386f8f14", + "ca2826a9ea444e4fbc8b2cd785dfc4a3", + "d8108f5dafb05591a0165983da77f78c", + "dd028d3ab7fb4aae84694d21ecf4c201", + "e430606893ee431087b189d71cd29ac5", + "f664534ad3bb423eae4c699ad1d1ad79" + ], + "train_samples_hashes": [ + "d3cc78c8149b05f11e9d0db3c823cb40", + "101e41519137ae547190419f525c4d94", + "cdd906c5b95a94a735bd54c97bd58a2a", + "6e8d970bba6c0c8952eda0ab9cd8cd72", + "08de11802046048c83d048556c02be24", + "72ae9881ceb2f8e55efa36fe0d6cb64c", + "ea935ea072de6c0dd32b104da965a06b", + "5eec7d64ce95c19cce60da50fbac57b1", + "e6161e86609d22c3e2b4a980d1aee886", + "2d260d089efc673bb4978b5b9318fc7f", + "67183565fd9cac0a84d61b80d05ca8f1", + "8828271b94284dc466380acbbabf6e83", + "70883e53be6c4efafeaa5db3677f2f9c", + "407143b8a9c3bffbb0212727c69f68e0", + "7ec41326259c36505116a76220602663" + ], + "test_samples_baselines": [ + "2dec7910f3a25f448d279e1bcd61c763", + "322aa53bc0ae43c08d03196c3fa01dcf", + "31dcf19a3b73436f9e844e7822ac52c3", + "b794f7c284485e548f68445b0c123a3a", + "22eb60ad8f495387903e93a80cf12294", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "4f105c22a575422faa161016192329d2", + "f3198e91ff7d593b8b1d00e99b154708", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "38482b8747bc4bfe88ac617356e9c6fe" + ], + "test_samples": [ + "2fc73742c4235516a326bc749d661972", + "322aa53bc0ae43c08d03196c3fa01dcf", + "3ad87b80437b43fda95e551a2efc7b87", + "906b2f75d36d5647a8fe4ef05f19fa5d", + "9646ae5511015044b647887e96eb8dca", + "aec60d90f7ef4979a80c56f5c5bdeafa", + "cd5b8bdef4d44b7d8f6682e3cde5fcf2", + "e2cc7f4083515d948057472541e749f9", + "e8e9190868644fc38aa1113756e04fee", + "ff3a19d5eaad46c1aed4ca4373c26de3" + ], + "test_samples_hashes": [ + "717393ab4afb9d948c8faa1abeb944d5", + "c2e3dade987a5c80aab3c25a1f5f0b21", + "6daa1de025af0c11ca69b85d58be04dc", + "1e8c2a2e489efb1d77918c94cd44b248", + "85c972828adf0db2de946c8143aa3955", + "7dfdef65098f35fb6f5804a59777d9cf", + "3087b5c9008db5fa7b442d4d329c21dc", + "7cd0b94e84b5ae80e347e423efa65420", + "5a5f4d277150617f1f739a6e55cc29e1", + "f27adfcf89db0298da43ee6f363db29e" + ], + "seed": 4, + "test_set_slug": "home", + "shot_slug_recipe": { + "is_adversarial": true, + "number_train_samples_per_class": 3, + "number_test_samples": 10 + }, + "accuracy_with_baselines": { + "euclidean_knn": 0.2, + "euclidean_centroid": 0.0, + "correlation_nn": 0.1 + }, + "train_test_sample_hash": "1b25cc44bc43c8c5810d0bc5678187baef568a8ab3697fc11159a5ae9dc1af21", + "other_samples": [ + "00af86611bf0452986e018425db6cdd2", + "03cf9a5fd2af43aa8b2413596e24ffa4", + "0655d91288195e93a7903500183734dd", + "0982071fbb684307913a380272a1ae91", + "0c5a280f084b4d1fa52dddd30269c1d4", + "0d79edc07243565d8d8b5380ef81e340", + "115036fcb4cb40278192aeaec3c948ba", + "1d716244585943cc9e58b841706c0a36", + "1dbde2fdb75b5ca294baa80857327b67", + "1f0a940c521643948cd6a4a3f8c15067", + "208ef509e8b04166ae5333a2877f1295", + "223c934b74a24e5f9df19997035eef2f", + "23cca9a4b30252e89b56e757b3b6b0f0", + "260750f9d27a4578ba899fcfb02ac653", + "2713905d0dfa4d56aac01b96531ab3e7", + "299fe66074144eeeae74735cf42d161b", + "2dec7910f3a25f448d279e1bcd61c763", + "2eaaaa5d6f8e4c1b8447b4389e52853e", + "2f00e9a90e3c4484aa48107c021745aa", + "337efe3f652b4c12b24b70a63f25adff", + "38482b8747bc4bfe88ac617356e9c6fe", + "38e9b9cde3d547a9924b3af04070203a", + "3ba3a08aac1441ab9f7062076ab2e050", + "3d97964fb7c24cc291fbc25dea19c9d4", + "3dc5e2e8d7f150fbaf80df25ba9c426a", + "3dee5963901d530e9f2fde3c84a55982", + "402307421702585c86a2ff40516deca2", + "421875a8a7f441c4bffd546a137058b9", + "450660a3c2d24f5bae3eeec80638199c", + "477fa4f8ca014f559465bdc2c2318c78", + "4936a5b652995d1cbb5987f3c7ee6866", + "4b91015695f94331ba68ab2f9bc8feb1", + "509ca665d1854fa596c901a5468197ab", + "52cb93e7e69d4abf85d6db293440f76c", + "52d91d5389a652baa478069ea293ca73", + "53d61728255b43158f043258bf085cb9", + "59b2dd9b949c42e7ab404612b867f051", + "5cff394d93ba45fc950eb525d975c078", + "633f8b94b4a64afab5c4183b67987092", + "660fdef92d435e0a88bb41b9e958b990", + "6ff497703dc54edb8534f212c5b7c567", + "7b689cd67c2d4021930e364b1c2ed8ca", + "7d1520b2be3c531ab4acbe645c9e18cd", + "825bb775cf154fdb983653ce20d329dc", + "833168b23c474c3f95c1ba4d9fff6176", + "8840791a44ca45f2b1e72ee59093cef4", + "8aa929de582d5cbf83aa53fe9ab0a174", + "8ffce32ebf95402db33ab77ea1fe5f12", + "924138e0d0f9479f8dccc89cc12e589c", + "95357a192b404d9d942b699a96f3f91d", + "96814fa562ac4bac8cdd1ace014be742", + "97c922e470a6420aaa0db38a7b47ab14", + "97f285a3809940f68129ed3a9d1d9b26", + "9b1982d86e3b5bb892c7d1270f5e8e4c", + "9e10b79464c24f36a53f252be6541d5f", + "9e2271b4625541f7964a03844c658511", + "a3dc646b3f7647079bfb49c9acaa18e6", + "a788bcbddb3e4143afc049ae6fd613ed", + "ab1dbdc0c2344be7a4b105dd445c3e62", + "afec7fdef34d4256a0060406c9b1cd7c", + "b6a00423859b48e8b634aaf7a4b4208a", + "b6e5860643684682b60e6c146121b2be", + "b794f7c284485e548f68445b0c123a3a", + "b7d2dbe41aef52c58ccf2a7fd13903a5", + "b9d5d04e374c47d58d820fca9d364a11", + "b9f73c1ba2e14456bc1cb2701b0d61d4", + "badd89b9580a4d319ec3f8a0bee814b1", + "baf1e1790569451993d611a8cb903120", + "bb5c824b74c65fd0bfdeeb1b5969e46c", + "bd084d19f2444af8a6f98ec7815d2882", + "bd1db72c3cdb481caa778ac3639cccb3", + "bd222ab0a4034903923ada8b7a09816f", + "bdf601681b8541659ad53f2b52d11526", + "c1e94ce33230421a847d36495497409d", + "c56def5d135e40c290812174a4d3a066", + "c5d7959f17da52dda612bcc328e0b427", + "cc910fa7368c5587804f005866a22079", + "cfcf113e827240a6b7e7a4be86078864", + "d1141599d62d4cd8a57708873e904583", + "d2354df65fff5ddf9e1b8e5d138282bc", + "d913f421c0964aea99408cec015c64c1", + "d950be4ceaf855e4b99175c9b68e37f5", + "dcc54ad4f3014c0cb2545e1f84a9ee38", + "e07a5edf78e844b8ba42f7b4add2cfb7", + "e0934b52a37759e3a90cd51c68f250e5", + "e40e7855b79058599803ffe83e69df2f", + "e4156c6755bc5b86ab4fb7c158deb515", + "e459f3032a4152a496af1745a8b75c10", + "e4a4fea8b64e496cb7e24f4de9f270b7", + "e4faa5a0e1ad4ca4b94bf2f18f4a093f", + "e6e67852460340c291d3b2d33978e955", + "e803fdb9e638552d9050b896808b9b46", + "ead1bd72fabc53758d90a08e1dcd2507", + "f2f13ebd94865bf28ae7eeea4ef29cdf", + "f3198e91ff7d593b8b1d00e99b154708", + "fb2367ca63f6468bbb95f4ddbd0d99c9", + "ffcb3e01741a516091f8da7aeb689274" + ], + "accuracy_with_baselines_all_samples": { + "euclidean_knn": 0.1, + "euclidean_centroid": 0.1, + "correlation_nn": 0.3 + } + } + ] +} \ No newline at end of file diff --git a/submissions/claude-4-opus-high.zip b/submissions/claude-4-opus-high.zip index b16a895d2213ab58cf9b74fd6f68d5cf743d74d5..e8480574aa542aa7858c054932504e238531d84a 100644 --- a/submissions/claude-4-opus-high.zip +++ b/submissions/claude-4-opus-high.zip @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7e17bada18899a0dc49101f353ad1514f3784faa313c63822cab785589d8fc00 +oid sha256:b3ff816a03c88a46c06e35e022d585e7d4c49eba45ea7efa8ccc90c5dc42c173 size 2656 diff --git a/submissions/gemini-3-1-pro-high.zip b/submissions/gemini-3-1-pro-high.zip index 0db3ab53aefc100c7a02302f631c30ce87a918ad..a85ab01020fc92ddff795b1c60e339f4737eece9 100644 --- a/submissions/gemini-3-1-pro-high.zip +++ b/submissions/gemini-3-1-pro-high.zip @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:5306e102472bdfd588bf5270ae7eb86edfbe9fa3a6bd1365c5545e5287d2aaad +oid sha256:21bc97a2bcf8436fc02123be0750a71488087ca3f37631309b0bfe915ed3f2aa size 4102 diff --git a/submissions/gpt-5-5-codex-high.zip b/submissions/gpt-5-5-codex-high.zip index 5fed4068b6288035bdbfa481747a872b317fbf13..3e7de950c4cf78168e1fbc14636f1290b000857d 100644 --- a/submissions/gpt-5-5-codex-high.zip +++ b/submissions/gpt-5-5-codex-high.zip @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7947f41bd53e93f9ed63642cc6ae63c240bfa1fd2ea2f15d084bb1902a025b04 +oid sha256:f43bdff6177bfa7473d7b7f69d06c43b6b65996510ebc348235c25385735aa12 size 4002 diff --git a/submissions/minimax-m2-7.zip b/submissions/minimax-m2-7.zip index 4aedcedf822ac657a6d5d010d81a105bd345270c..ce5759a4a34a31d8dfe2553b6b7c2d02c98f10a9 100644 --- a/submissions/minimax-m2-7.zip +++ b/submissions/minimax-m2-7.zip @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f53c8c9ca4eea34cac11e9bbaf6450fca6f7e7cbc4e71058fc827d11cd1446f0 +oid sha256:6421e235e321a59774d81f6425bd02944f1fc679ed443b18646a7108fb7e1061 size 2822 diff --git a/website/environments/BallDrop/description.json b/website/environments/BallDrop/description.json index 9f8dd40908a4120c9f19448ab50360cb5586758c..a48d7869f6867df80e7357d29c864ea4fdaecf03 100644 --- a/website/environments/BallDrop/description.json +++ b/website/environments/BallDrop/description.json @@ -5,7 +5,7 @@ "coefficient of restitution", "gravity acceleration" ], - "download_link": "https://huggingface.co/datasets/TommasoBendinelli/tsenv-benchmark/resolve/main/questions/BallDrop.zip", + "download_link": "https://huggingface.co/datasets/TommasoBendinelli/tsenv-benchmark/tree/main/questions/BallDrop", "environment_id": "BallDrop", "name": "BallDrop", "observed_channels": [ diff --git a/website/environments/BounceBall/description.json b/website/environments/BounceBall/description.json index 80d6c777edbc76477d815b27d203f16910a96226..d32d675bea2bfbd6d9ff8472f9b166f8fa214d8b 100644 --- a/website/environments/BounceBall/description.json +++ b/website/environments/BounceBall/description.json @@ -6,7 +6,7 @@ "coefficient of restitution of the left wall", "inclination angle of the rail" ], - "download_link": "https://huggingface.co/datasets/TommasoBendinelli/tsenv-benchmark/resolve/main/questions/BounceBall.zip", + "download_link": "https://huggingface.co/datasets/TommasoBendinelli/tsenv-benchmark/tree/main/questions/BounceBall", "environment_id": "BounceBall", "name": "BounceBall", "observed_channels": [ diff --git a/website/environments/MassSlide/description.json b/website/environments/MassSlide/description.json index d0b523396a041445e43b0f8ae6342362de99295a..4ece001a0f0383a2bd2868a55231b53ce8eb484a 100644 --- a/website/environments/MassSlide/description.json +++ b/website/environments/MassSlide/description.json @@ -5,7 +5,7 @@ "Coulomb friction coefficient", "breakaway friction coefficient" ], - "download_link": "https://huggingface.co/datasets/TommasoBendinelli/tsenv-benchmark/resolve/main/questions/MassSlide.zip", + "download_link": "https://huggingface.co/datasets/TommasoBendinelli/tsenv-benchmark/tree/main/questions/MassSlide", "environment_id": "MassSlide", "name": "MassSlide", "observed_channels": [ diff --git a/website/leaderboard.json b/website/leaderboard.json index 403254c8085c2c6ad66939f11dfa6762a5d4bdb8..8b1f5b70d44fc0c2a775bbd6a46407fc66721d61 100644 --- a/website/leaderboard.json +++ b/website/leaderboard.json @@ -25,7 +25,7 @@ "Direct" ] }, - "generated_at": "2026-06-05T14:56:45Z", + "generated_at": "2026-06-05T17:59:03Z", "rows": [ { "agent": "gpt-5.5",