| { |
| "checkpoint_step": 100, |
| "checkpoints": { |
| "binary-mean-center": "7d5ffe3cf730f06f89cca32182896ef1570977bae7795fac5df4680d349985ab", |
| "threadweaver": "d937f4406717b5925fc3127576a60ca968f41a8f027f358b4e3389190b7e3893" |
| }, |
| "runs": [ |
| { |
| "run": "original", |
| "repeat": 0, |
| "base_seed": 20261015 |
| }, |
| { |
| "run": "repeat_1", |
| "repeat": 1, |
| "base_seed": 21261015 |
| }, |
| { |
| "run": "repeat_2", |
| "repeat": 2, |
| "base_seed": 22261015 |
| }, |
| { |
| "run": "repeat_3", |
| "repeat": 3, |
| "base_seed": 23261015 |
| } |
| ], |
| "protocol": { |
| "source_sha256": { |
| "src/math_recursive/constrained_decoding.py": "88f11c540f17d736f232a2ac6bc4b4f7605f94ec8a5da244d37a3e354d99eebe", |
| "src/math_recursive/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", |
| "src/math_recursive/tunix_recursive.py": "e507ba005391b8359cd2cfdd59e1431ad1eebb93c16e143da59c15317e71f74b", |
| "src/math_recursive/tunix_replicated.py": "b892c20b09ca6e8637ce8f2974504ea84829b85ec6932602b569f194adbeb4d2", |
| "src/math_recursive/json_batched_inference.py": "43fe5462aceeadfeeb96227c33bd931dde0d38d63c3c04b47d312c87e57071f3", |
| "src/math_recursive/tunix_semantics.py": "96e814dd8d9a3b4435bc0a02c282f8f0b9350d3e25c67f9f43369f40cf4ab9c0", |
| "src/math_recursive/tunix_reward.py": "1daa974afdbce78687e549b6738c04f043363d93af939d21636e9307808f8162", |
| "src/math_recursive/tunix_lora.py": "371d87724dfcc94db92f0f3bb70a2fa3e1bc625823d4f7037d2add4be78478cc", |
| "src/math_recursive/checkpoint_storage.py": "035e9051e59aa1b5a120449e1964f85deef0394aecdd2149ca2426e827aa51b5", |
| "src/math_recursive/grpo.py": "2b060bfb1982aab39665934c5489655989b8cafd46c2e33c3fb643fd2285ce0b", |
| "src/math_recursive/constrained_grpo.py": "f4437189d14862aa9058d5e60813ff4676481d67ae588866928fb9ebb7b80754", |
| "src/math_recursive/tunix_multiworker.py": "ac9f57475c6336bd2ee6255e8cb5755807f7dea48c3eea7818042fb1b9615c76", |
| "src/math_recursive/tunix_memory.py": "fd27c7bb63642e9a425a9aa1180ce8045325bb154a5bde86a930055678f40373", |
| "src/math_recursive/call_grpo.py": "b84e4551718246b7c66fb1e52a8fba5914938c3de85aa7eccd69faf7ba708b69", |
| "src/math_recursive/json_inference.py": "1b8253fe1ac0ceb3389c49c8e4ad4d7cefaa720c1886e4e7148adad35373cf3c", |
| "src/math_recursive/engine.py": "4490efd722b142dc7ec8d0e759f45136c612e88fead0770528ee60e1d127fc5b", |
| "src/math_recursive/flattened.py": "60dd6dab1b408c67da85a5c014aee819d45e39e8b73e0e532ad5eef0869670f0", |
| "src/math_recursive/evaluation.py": "cb8a03b3fb730ecbf06fcfd224556b967bf2218bd91a69e97d6c5fada5c3d32a", |
| "src/math_recursive/forced_constrained.py": "267aaa06fef89096c141460e8a587970822bdcaeb44c6acc2e16da93772087de", |
| "src/math_recursive/json_protocol.py": "cc686d4e5f0b345ab4c93ab2ad741656456d40e8fa37ff3f84430e36f4ee7027", |
| "src/math_recursive/tunix_checkpoint.py": "4f0c1fdf485b214a5ee01dc4c239bcdba45f19683bd83658cbbf1cccb86b2266", |
| "src/math_recursive/tunix_metrics.py": "2e409246c83fd147768458f9def4b58ad1b896efe8216f03f21599d19a2027b8", |
| "src/math_recursive/json_fast_constraints.py": "bfec2baa6276321a4b65af2722933358ade49903a390fcc8d6afd9a9114b9440", |
| "scripts/evaluation/check_recursive_json_feasibility.py": "6420b2e5e1b85be5a29bfc769cbe547761f4f00c7aec474f301ef752cebf7d31", |
| "scripts/evaluation/compare_metrics.py": "02ea10f8be46a2bd5da016ac9d6457267f7f3459afa903c972d7eaeb1683237a", |
| "scripts/evaluation/run_sft_baseline_gsm8k.py": "dd9b38d70c38b87cf2101b1b74d60ea1d6b615c64bc52d2e50473922b590f4d3", |
| "scripts/evaluation/aggregate_tunix_root_ablation.py": "fe8df89745417d07f00c92dac2131eb019ee9d00fe1a8ee2ae33d8b8a4e65201", |
| "scripts/evaluation/run_flattened_eval.py": "ce5cab227c13b99007b2002b82076feec865f5f6ae67558832264a12b0c759f7", |
| "scripts/evaluation/audit_sft_baseline_rational_comparison.py": "4210910e8fd0ce18bddb6a74023620b8b84d4f14a694c42f7050bb301694e960", |
| "scripts/evaluation/audit_v31_rational_false_negatives.py": "9b10ced9f3898dd58cb515f6f3664dfcfb2e1da49da982883f22241a476eaa03", |
| "scripts/evaluation/audit_flattened_vs_recursive.py": "9fe4bf9aa753728906125684f35c8e61bb0a30336bf343669234d213b5d9c87a", |
| "scripts/evaluation/resilient_explicit_grader.py": "d057a86cf76b27efa302ddaa22402c2ee88bb57c0f218b2c7d9a5168289554af", |
| "scripts/evaluation/report_call_format.py": "330c8c0a506b426b0f45b1c6b70d347cbbb7dcf0d461e74f9af9f9042a224f19", |
| "scripts/evaluation/aggregate_flattened_eval.py": "142e6486379bd274e4000729e33ec3ae5ae91d4d68cf5e53e7d8d54985eb7409", |
| "scripts/evaluation/run_tunix_root_ablation.py": "19c59919b4ff3648c4a1037af03b7ad0122764a0dbe76fd9e946dbc934905679", |
| "scripts/evaluation/eval_call_grpo.py": "9c3291a837ba17c85b1bac45d2c32a90dff9ea82f45ac8484ef441a523f66e89", |
| "scripts/evaluation/grade_explicit_answers.py": "cf736934c347d71028c27589eda95939fb460c80407090a5e6ceb611b13e2004", |
| "scripts/evaluation/run_recursive_eval.py": "2123bc1b9042d1c7bfcd9bd35f8c95556b1692c65fc2d0b8541433d6a460cd1b", |
| "scripts/evaluation/validate_json_batching_gpu.py": "f5ddfb57bf1c47874180d047d27d4b4ea6198437f8a424d8f0cbb6a8b7015e7b", |
| "scripts/evaluation/run_constrained_grpo_eval.py": "06e9cf1ab502868f066cec63066a240bd76650553f4b216605feac24f6930fc2", |
| "scripts/evaluation/audit_v31_math_format_failures.py": "55b1e21120b18517a09cdc0318ece84d2b4afe9d34c4578476e120846055dd23", |
| "scripts/evaluation/prepare_tunix_benchmarks.py": "6a52f22f8241ea86a318c9eaa60fe0c5b98ade333c3a7caae88e08a71d85d376", |
| "scripts/evaluation/run_tunix_eval.py": "dbc370124025ce7f1c3792fa282fac22a0e35baae28257c017d2d80990845637", |
| "scripts/evaluation/validate_json_decoder_gpu.py": "f12b36d39268d8c21ca70dbe7481e0b064d0a6a2cc06987e2c8fa2633f260e14", |
| "scripts/evaluation/summarize_rollouts.py": "e1c79d6ee7f0c5cc85b0360c8b3103329c2a6d3bc024000ce652d99784c2abd3", |
| "scripts/evaluation/audit_gsm8k_baseline_positives.py": "c80687f6547447e9f88392730434da77e6398a16bf65cd925469337f0a612e89", |
| "scripts/evaluation/profile_json_constraints.py": "b27cced0ae56dd3bbb7a2f27b4bd5f5f7ba05ea3b04a2838435d864539f6b275", |
| "scripts/evaluation/aggregate_tunix_eval.py": "f55caf489905885aca4874bbd6c90922cdf1361bcfb893e3b9fc7fa3c0d18993", |
| "scripts/evaluation/run_qa_eval.py": "d24b0be95699c121875588eaf808bb2797cbceeddcb96cc8faf9e8bb60bd5e55", |
| "prompts/recursive_sft_prefix_prompt.txt": "dc622e5bc70a1a5970827d70ca764128b07a9b32514ebe2c032ce9da7a147f85" |
| }, |
| "versions": { |
| "jax": "0.11.0", |
| "jaxlib": "0.11.0", |
| "flax": "0.12.8", |
| "torch": "2.10.0", |
| "google-tunix": "0.1.8", |
| "vllm-tpu": "0.31.0", |
| "tpu-inference": "0.31.0", |
| "xgrammar": "0.2.8" |
| }, |
| "call_batch": 256, |
| "active_trees": 128, |
| "engine_sequences": 32 |
| }, |
| "reported_three_repeat_mean_excludes_original": true |
| } |
|
|
