Update coding SFT holdout coverage and provenance

evaluations/mbppplus_n160_t0.6/README.md

@@ -0,0 +1,3 @@
+# Llama coding SFT checkpoint evaluations
+
+Two newly trained Llama ordered/shuffled baselines, ten checkpoints each. Canonical MBPP+ 378-problem holdout, n=160, temperature 0.6, top_p 0.95, 3072 response tokens, 1024 prompt tokens, prior execution-based scorer. Pass@1/2/4/8/16/32/64/128 independently recomputed from raw binary scores. Final model weights match the step65 checkpoints. Complete coverage requires 20 evaluations and 1,209,600 scored completions. This extends the 13 previously completed coding SFT arms in RL-Forgetting-Experiments-3/mbpp-code-sft-artifacts; it is separate from the coding RL training-set n64 temperature1.0 results. Partial coverage is not completion.

evaluations/mbppplus_n160_t0.6/delivery_manifest.json

@@ -0,0 +1,8 @@
+{
+  "status": "partial",
+  "expected_evaluations": 20,
+  "expected_completions": 1209600,
+  "n": 160,
+  "temperature": 0.6,
+  "units": {}
+}

evaluations/mbppplus_n160_t0.6/provenance/arms.json

@@ -0,0 +1,182 @@
+[
+  {
+    "name": "eval-llama3b-ordered-s6",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 6,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s13",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 13,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s20",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 20,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s26",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 26,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s32",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 32,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s39",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 39,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s46",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 46,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s52",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 52,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s58",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 58,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-ordered-s65",
+    "group": "llama3b_sft_ordered",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "step": 65,
+    "source_job": "llama3b-code-ordered",
+    "source_id": "sad_drawer_63s9dw41sd",
+    "result_root": "7206895844.05355-a82cdfde-bb79"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s6",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 6,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s13",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 13,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s20",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 20,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s26",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 26,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s32",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 32,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s39",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 39,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s46",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 46,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s52",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 52,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s58",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 58,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  },
+  {
+    "name": "eval-llama3b-shuffled-s65",
+    "group": "llama3b_sft_shuffled",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "step": 65,
+    "source_job": "llama3b-code-shuffled",
+    "source_id": "willing_napa_b29yxdlm8n",
+    "result_root": "7206895844.05268-768ae95e-2970"
+  }
+]

evaluations/mbppplus_n160_t0.6/provenance/source-hashes.json

@@ -0,0 +1,399 @@
+{
+  "run.sh": "601d2b3ef8aad421bc10fe39a067f2c4b717cedde4e63d40ca3aa33bc7d30cc8",
+  "run.py": "ae8d247389e2207d6b6b867df4634e6d4a198ad69ebd0e3f64bf1e7b6e8293d9",
+  "reward-signal-analysis/hf_login.sh": "6df01457fd3a0dcd66519e82bfd094b34652170b0fb812ce705c787e79fa60fa",
+  "reward-signal-analysis/hf_model_download.sh": "708fe290ffb65e0f09b96d5e35e9df592ff52fd40b8178c5ffc37643a3560cac",
+  "reward-signal-analysis/reward_function.py": "5b3028650885a539cff15031cec802a0f1baf4acaf87985cdc4b3b8a2a12dc4d",
+  "reward-signal-analysis/setup.py": "7c572aeb0fb969bf1373cf00f6ddacc6f678ba8d0c7b58f85ca3562108310659",
+  "reward-signal-analysis/code_reward.py": "9a3924e46d8d5acdde0a8755dba21a6cc4b641c4e04426b8f87ee47dd8d262cc",
+  "reward-signal-analysis/reward_function_agent_prm.py": "a73514a6b27e2e1631a53d2cc0c7d1fc5b32a4a827678c6b60f79cd4523ea869",
+  "reward-signal-analysis/reward_signal_utilis/split_arrow_files_into_4.py": "c76f8c56385b5d356cc11639097b2e74fd552dc500b4947f72b7a9c970f971d1",
+  "reward-signal-analysis/reward_signal_utilis/data_view.py": "212ffc38eea150ff8774c5132502ccbd0db2b500a6ab5c65d2069a8aea813475",
+  "reward-signal-analysis/reward_signal_utilis/download_Nemotron-CC-Math-v1.py": "23da1515bd2f0a294e46a6e3486e9463da7983ba289ea405146724b1c063a881",
+  "reward-signal-analysis/scripts/run_math_nobuf_kl0p001.sh": "fae570579b5b40a9af2c101ea2af8556ca0d752b6af15d5271f9e88331f1b814",
+  "reward-signal-analysis/scripts/run_math_nobuf_kl0p01.sh": "6593636e3ed5dafec1c049808c65b45bfb0aa7ad9a9f0abb7536991d896ca344",
+  "reward-signal-analysis/scripts/run_kk_bs256_strategy.sh": "7f93b0d9af45b2aef5bc614ed43e2e592a9e34063020cf96eb48bdd0c9b8a3ae",
+  "reward-signal-analysis/scripts/run_math_hist.sh": "52c7142273a4e098306efa4c1e2db8a034a68823cdccf5b052abca9b6583932e",
+  "reward-signal-analysis/verl/protocol.py": "61f24d96ac3f3bc808120afac76d43bbde7bb51044962bb7bb8623e9b989563a",
+  "reward-signal-analysis/verl/__init__.py": "d8dfaa3f43a410a71832b8dcb3fa0c5c9ed5332e78e3b34a52539032b919666d",
+  "reward-signal-analysis/script/dump_math_anchor_rollouts.py": "491a366a28510a6e709f1fe2b5b445535f19c2e179ba7b900bd53571d12b034e",
+  "reward-signal-analysis/script/eval_math_pass_at_k.py": "c773b99e64a863f29e587413965377e39100d73caeed0a1797809378e244a9ab",
+  "reward-signal-analysis/script/convert_checkpoints_to_hf.sh": "a552fbb3c3cb8967c83e7b62e68a012fdabb4412ac95a115e22d30fd8a8cd0f8",
+  "reward-signal-analysis/script/submit_rl_eval_kk.sh": "81f0030677ac9319901ae14489c65558fc2aa43297e9131d14d4b5241b1a1add",
+  "reward-signal-analysis/script/pass_k_eval.py": "588ac3a42335be5ab6c5d639da81988a44edcc00895aed564476d1460c1eafb6",
+  "reward-signal-analysis/script/aggregate_results.py": "43c81b93234d1edf7a54e47b198769beb715bd464a4d9afbc0bf09b6cecfe2e8",
+  "reward-signal-analysis/script/run_math_incremental_eval.sh": "fa22326e1eea5c5b38ffce491da5a4d55e75116115a5e3485b184b25f1269f71",
+  "reward-signal-analysis/script/eval_math_loss.py": "5faf097add8d0cd6685ded694f44a18501556622c4b2c4118c8fc580c21baf0a",
+  "reward-signal-analysis/script/retry_failed_eval_jobs.py": "a7e6038a9c7fe953a63c8364c1fce3acfbda70cd11f81c0d44ddd82867b3cf45",
+  "reward-signal-analysis/script/reshard_math500.py": "047fce106dd958918f4c6d141a684a9f21688da9f24fc822d5dc564d0d473e19",
+  "reward-signal-analysis/script/make_deepscaler_eval_yamls.py": "2faa25314c02eb09c1447d01de1436619ff1e49d9cd8545fdb25c8e9d557a6f6",
+  "reward-signal-analysis/script/run_evals_math.sh": "9840ebcbf312f2dc3d8d9d29763b7bd421ec1ddd3b94ee7241d4861645c87ac7",
+  "reward-signal-analysis/script/verl_eval.sh": "45ad7ae016ce08225e6389cdf61ddcc0c7a15f48de5e0669b410eb23ba9df9de",
+  "reward-signal-analysis/script/plot_deepscaler_passk.py": "11e2e7443bb4341c4bc9f6b51f491b6724af47fa43e2e9ba636c62d2173c3e23",
+  "reward-signal-analysis/verl/third_party/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/tools/geo3k_tool.py": "74052601c0eb006d3c07cd16c336219c91aba2b8494b644aae321557391b6ac6",
+  "reward-signal-analysis/verl/tools/sandbox_fusion_tools.py": "87ef0306060a64a507cf2faa36f2cd36885e8d48a18fa66638848a0f071c649e",
+  "reward-signal-analysis/verl/tools/base_tool.py": "7d4921f6bf175de9e8f7581cf5b7aa79ebf6275a124e02d6f6c83735ce0d1cf6",
+  "reward-signal-analysis/verl/tools/search_tool.py": "b42caa4b07c3e13c674ce0e7f0911fdd49ff3b34fbd2c4064a9a683fa39b5086",
+  "reward-signal-analysis/verl/tools/schemas.py": "aab51d994e2353212e170db0870e360c4aa8b1700076a8a571203a153b97233b",
+  "reward-signal-analysis/verl/tools/mcp_base_tool.py": "4ea99f289c064bfe55e14c8207b0e99eea1435995eb13d9ed1fbd45235d602d6",
+  "reward-signal-analysis/verl/tools/mcp_search_tool.py": "729afdf2b77e8f5b2fcca990e2b2908ecc1da0d4a4b1b9fc79a339e2c790362c",
+  "reward-signal-analysis/verl/tools/__init__.py": "0bae18fab65cb4028732b91a1215baf8e02f9fdf8b4762a523662a34374e7432",
+  "reward-signal-analysis/verl/tools/gsm8k_tool.py": "c3e7e3bb56b66b2331119137c94e4aade40e70366f72b650be2f45a293f987e2",
+  "reward-signal-analysis/verl/workers/fsdp_workers.py": "636b7cf66a803decd1dc81d7852d6355c52caac8f030ccd1f692fcb3ae03499e",
+  "reward-signal-analysis/verl/workers/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/workers/megatron_workers.py": "87f371457a794c1cdde143c6ac97976d57393d4346084221ff3b5f9c4a5cfca2",
+  "reward-signal-analysis/verl/single_controller/__init__.py": "bc8a469c28cee85f4c7e7bcf75733a03316a56adfb8254abd33b368253b07fe3",
+  "reward-signal-analysis/verl/experimental/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/trainer/fsdp_sft_trainer.py": "3214b9f75ee0e578f0777b5aac44308cc1c26cb7534a3120e4b084f37c9f88b6",
+  "reward-signal-analysis/verl/trainer/main_ppo.py": "0750425ff374ed0f9fb1aaeb095a12a5ed4fc06258b2ba96babf1e77cf452d71",
+  "reward-signal-analysis/verl/trainer/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/trainer/main_eval.py": "59cd305bccbd08061c54a218abdbdc086d60fabaded2022842af6f50b8ca88b8",
+  "reward-signal-analysis/verl/trainer/main_generation.py": "edbcf4481879ead51169f64dab44dad63176a9ff9ab117e18230fb736a8fa8e0",
+  "reward-signal-analysis/verl/interactions/gsm8k_interaction.py": "87f0bb745ddcf7ea5e3b7717c5281b08cfe261e14801416606070f83c9847398",
+  "reward-signal-analysis/verl/interactions/__init__.py": "48e4a0c1591de920175cf07b36cb6f87e094becf910af84c9db2d527a9464497",
+  "reward-signal-analysis/verl/interactions/base.py": "3ea1836933ca0d36adad6414abd40fa2e0b0ac1a7684b9c00e3cab8f5cbb0be4",
+  "reward-signal-analysis/verl/model_merger/megatron_model_merger.py": "3098ce54d11d6c0a6d6c2ab360d3cd77fdbbd7a52e12bc60a2da8fd123bdb676",
+  "reward-signal-analysis/verl/model_merger/__main__.py": "d5a06b42997fb9e5cc229da996262f0afa26c39d86aba898af9eab562ae6a8ce",
+  "reward-signal-analysis/verl/model_merger/fsdp_model_merger.py": "2ae5f3f7fc95871b08a2e185def4f1a5cec09164579cd15eeb2314cb73ef1f0f",
+  "reward-signal-analysis/verl/model_merger/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/model_merger/base_model_merger.py": "f644e942aa9c782ea0900b41ff70ddb45cfb550651c8bdc8549f73865f89dadb",
+  "reward-signal-analysis/verl/utils/fs.py": "e3cbe6e1d000ecf7ed7aa59f5b0a694a04e1413606887eb81c4ee6e5e5518f65",
+  "reward-signal-analysis/verl/utils/model.py": "8bb8da222fd646c8391c239a18f209f26a53f7c941b3f05bf367eacda87abde4",
+  "reward-signal-analysis/verl/utils/torch_functional.py": "a871c79f2f4a9348a597848a33fc8e5e2e27c156f32b221696d9be9adb67f1f6",
+  "reward-signal-analysis/verl/utils/device.py": "6bb9bde1f330593647a94ba6184a5e606c1ae12c0a8e91571510edfbf5d2d101",
+  "reward-signal-analysis/verl/utils/tokenizer.py": "b81a09a2ac5a4e8fd3df37712b19d336d4174f40903b76ab817453282fab6205",
+  "reward-signal-analysis/verl/utils/ulysses.py": "51fe8a22602d6060188fa29ff037368d26cebf5e6435781e6bfc7d3b6bd20dfb",
+  "reward-signal-analysis/verl/utils/import_utils.py": "8ad791d8b9f2f0cdf6302e05b51fab8885a78a455cf4309da38d2970ddb3802e",
+  "reward-signal-analysis/verl/utils/hist_baseline.py": "f5ba0c55aedfa7c9fd5cc595248d206395f714fa53eaaf69e23ce8a07175078b",
+  "reward-signal-analysis/verl/utils/net_utils.py": "3088a0ee3f514390cf503abbd53e6b0d68369383bd9585e566d476033bd9d5be",
+  "reward-signal-analysis/verl/utils/activation_offload.py": "fb56a29a3f623b364470184ff481ef34e1e1205874175e1fa25dfee1d791ecb9",
+  "reward-signal-analysis/verl/utils/logging_utils.py": "72f6abe76274ecee5812c90879dc8f0b40e68d105c91da32a8923b8862320294",
+  "reward-signal-analysis/verl/utils/distributed.py": "e50e6d91818c80564b05a7c8e298fb0e554092b6efabdfa2d87886856a8813b5",
+  "reward-signal-analysis/verl/utils/torch_dtypes.py": "fc10a925f0ff4c1d103daf4643e436892f76ec00c02d930414ec9c2bf7a05aaf",
+  "reward-signal-analysis/verl/utils/tracking.py": "e91a7aded1df3cccc6fc9ea2cacb3fe18f90b9f5e89dc5a9c1734d1138494332",
+  "reward-signal-analysis/verl/utils/seqlen_balancing.py": "2d3a92ff748ddb9cced49a4988b321ef758c2a08e7646a7f9d91bb4e8c243d69",
+  "reward-signal-analysis/verl/utils/vllm_utils.py": "269027b94721a51b3fb0f08ef1a187c597400cffd79e21d8648b9adfc6eb5af0",
+  "reward-signal-analysis/verl/utils/ray_utils.py": "3f973ecd1ef61bfa3a865206a0e20d92083a5818b94fb83d02908f6c0d473273",
+  "reward-signal-analysis/verl/utils/transformers_compat.py": "6538eea7babcc780ff89f2f3ca5a9fa1848cf66218548fadb441137b96b8d0d4",
+  "reward-signal-analysis/verl/utils/flops_counter.py": "a9d6a49dbd79781f43b281ebdb3dd7ef6682dc795e58c3ccbbc3e03cd3495516",
+  "reward-signal-analysis/verl/utils/megatron_utils.py": "17ad3f746fa158f8a7a6f960bea4b8a5869f9fd1982bd859f8730273bcb754ba",
+  "reward-signal-analysis/verl/utils/fsdp_utils.py": "926d1424bc9802e4548eba8d9454c7ce22b63f0df1003815564b7e8851483bd1",
+  "reward-signal-analysis/verl/utils/py_functional.py": "df578722b06f4652ec1fa09917419a24836c54c80d686b8604626d49fb919659",
+  "reward-signal-analysis/verl/utils/config.py": "cd4e5b514c2e9bb37a756724e9789684825973d2ff6f6418bc5dc0a386503f1c",
+  "reward-signal-analysis/verl/utils/memory_buffer.py": "f96c00b5e5e5e62360c55088070a695ad3cdc9a6e3560948f07798af57dc159c",
+  "reward-signal-analysis/verl/utils/__init__.py": "491e8b6c6aa19b06dafcb72e706b1a30e79783d302f7909d3032ae699da8ba9e",
+  "reward-signal-analysis/verl/utils/hdfs_io.py": "14ea05cfdf000666899109a417d8d8bd23cb02afc43e06c16a861a3020590bd8",
+  "reward-signal-analysis/verl/utils/replay_buffer.py": "d75825d540467bcf66562870d5b61a3230f187011f1f0d92e6dc6cdb0a81c27c",
+  "reward-signal-analysis/verl/models/weight_loader_registry.py": "889d5d7ad4c68144c2d0e0b9cfc167c945baebdc69b83df36df0dbecf8312a03",
+  "reward-signal-analysis/verl/models/registry.py": "7c1523431efbf1977b583c7f8359ff2f2cfbb9d3c4af6cdc9f492f6c9d0370f0",
+  "reward-signal-analysis/verl/models/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/third_party/sglang/parallel_state.py": "254a9a6c9523e5794ad400e9842de52b1b742e7b3733f394f1703c3d96e158ca",
+  "reward-signal-analysis/verl/third_party/sglang/__init__.py": "566ce76a0185427357ae049e2c5e346ba6f73fedc32690f30862b3cf275bd7dc",
+  "reward-signal-analysis/verl/third_party/vllm/__init__.py": "40fa70ea891a300f3ca82e88efbff625efb5dc18ce4b092d376d1133f50d3779",
+  "reward-signal-analysis/verl/tools/utils/tool_registry.py": "c2cc1f97d158897ca54f8b06a7c0d42f198c1c34103927e94a40a49f6b4c1fc6",
+  "reward-signal-analysis/verl/tools/utils/search_r1_like_utils.py": "b9d799d1fd6c1b03d8bc9f350d29cb48f2a7a03d7676d37da80fe384b1903b61",
+  "reward-signal-analysis/verl/tools/utils/__init__.py": "0bae18fab65cb4028732b91a1215baf8e02f9fdf8b4762a523662a34374e7432",
+  "reward-signal-analysis/verl/tools/utils/mcp_clients/utils.py": "f8b0f14030e90d5fce8e58d6beaf301b5477e1d2d40dc56b5c41cacc0a7429c0",
+  "reward-signal-analysis/verl/tools/utils/mcp_clients/McpClientManager.py": "63da1edd39965a601021e79e97bd2165291f92aa04fe2b489f149e334a760c37",
+  "reward-signal-analysis/verl/workers/critic/megatron_critic.py": "5734ed8589d131bbadbaa64a756ded258468e0ec06d7f6b63d063e697473c4f7",
+  "reward-signal-analysis/verl/workers/critic/dp_critic.py": "cba2372b0b8ec8f6ddf07143c46b8d8263a517647c9f02ed17e69490e2bb93a6",
+  "reward-signal-analysis/verl/workers/critic/__init__.py": "a61c624ef2dd6bec59cc59302d88d1ccdc151ea5beb776b9733ac19bcfbe8fb7",
+  "reward-signal-analysis/verl/workers/critic/base.py": "9088e17a21d2aea6c524380e1eeb10eaf19df2719cc5d343cb4debafd8da033a",
+  "reward-signal-analysis/verl/workers/actor/dp_actor.py": "e1f951b0a8a5d6f8b02c9216de721cd0c18d4ab58068f1b98041bd04e5f77d65",
+  "reward-signal-analysis/verl/workers/actor/megatron_actor.py": "9399a97c18eafe4209858d777325c190d056360020c983470d606a14af7b013a",
+  "reward-signal-analysis/verl/workers/actor/__init__.py": "3053474474716dc57f7e1b5d80d23b28c7b146d9257f8b6bf3b7878de4eae219",
+  "reward-signal-analysis/verl/workers/actor/base.py": "fe023ef5dece8e915a4b38e15e06daa85381daae569b963f38a92f0a834907cd",
+  "reward-signal-analysis/verl/workers/sharding_manager/megatron_vllm.py": "7538a6db2ab97fe17cf74bbe95ca709a5dc48acde1b4a82be58775c416af6512",
+  "reward-signal-analysis/verl/workers/sharding_manager/fsdp_ulysses.py": "196bddf4f25815a48f94f7e8991f1dd8bc1477204acf6e0b0ea54a553a0784dc",
+  "reward-signal-analysis/verl/workers/sharding_manager/fsdp_sglang.py": "0200b70ece12bdc495318f2ee4c065340020325342192c1d51f7731e6fa15710",
+  "reward-signal-analysis/verl/workers/sharding_manager/megatron_sglang.py": "f35dc52b5fc69a93e4bab32fc8d3fdeda7dbe17dddd0ca9fccd5be763c13de1c",
+  "reward-signal-analysis/verl/workers/sharding_manager/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/workers/sharding_manager/base.py": "23a713b4f06336e6768acb4bf8621e05256be32af3abf2f0bd8accc10082a974",
+  "reward-signal-analysis/verl/workers/sharding_manager/fsdp_vllm.py": "e1204549388f04853e89a8baffc608c27b9fa4fd5cfa95eb332fb67e822e637d",
+  "reward-signal-analysis/verl/workers/rollout/hf_rollout.py": "5dc5f347ebbce3fe26d77b45aa106824ee0d81befe186646a03581eadcff01aa",
+  "reward-signal-analysis/verl/workers/rollout/chat_scheduler.py": "ae319f397ecbd79ff9a4c4476b33717792975d66b5c66cd27ff08b1410a9f8ba",
+  "reward-signal-analysis/verl/workers/rollout/tokenizer.py": "92b3315e760b62b5c6646c6918d31a7efdc668fa9edd43ea4f48d3f1f1ecba1a",
+  "reward-signal-analysis/verl/workers/rollout/async_server.py": "f3ac635907e55110b9b63aca2433a37c5584568c6b927f85044967b625fdcc14",
+  "reward-signal-analysis/verl/workers/rollout/schemas.py": "169d5548206c754e5d4d3cea77f993c594590573ee06ce97dd75e328b8c688aa",
+  "reward-signal-analysis/verl/workers/rollout/__init__.py": "4e772ac911483c3fa4ad7b301f532d10fd9a10ccd6fcc0aaa82a29cd9df14b7f",
+  "reward-signal-analysis/verl/workers/rollout/base.py": "962d7ba51e86e72da1a8254e4ff39cc9952bde8db85dee6507bdf3f6a279ad5f",
+  "reward-signal-analysis/verl/workers/reward_model/__init__.py": "c8e1a52f1a40f97bb885a5667ce0bc936f09d8416ca90243148417f416c4456c",
+  "reward-signal-analysis/verl/workers/reward_model/base.py": "9e3c68d62138dc39ca99b8f1cbbe01c24cf9b3a85a6a30e08fe59c910452de18",
+  "reward-signal-analysis/verl/workers/reward_manager/batch.py": "dbdcc39f4443f60eadebbc9130dd9cefce2f9da73410f4bc8e554fdb16932bd5",
+  "reward-signal-analysis/verl/workers/reward_manager/registry.py": "4178c70c6317a8a18103eaabedb167bd63cd1fd896c8271c5ebd1e552e7b49a9",
+  "reward-signal-analysis/verl/workers/reward_manager/batch_copy.py": "bb1628f83aefd7966492a2f187d71f417b4266a90af1fdaff1228b7f023faa9d",
+  "reward-signal-analysis/verl/workers/reward_manager/dapo.py": "342b3ca431734b9911cb0ec14be0f32261fb284d53cc6a4e8c7f71517dcf9cf5",
+  "reward-signal-analysis/verl/workers/reward_manager/yr_code.py": "2eaf369b044706e115e5c75e182f52e37a0ea9458c194ddaffc420c38b636e81",
+  "reward-signal-analysis/verl/workers/reward_manager/prime.py": "f684370014322bc26899744bfa9300f8f4c1ac7c9b4867c901775065aba647e9",
+  "reward-signal-analysis/verl/workers/reward_manager/__init__.py": "fffdb47c1e90965c97410e77450cb299a6f5521b403484c6f2d341fafd42cc35",
+  "reward-signal-analysis/verl/workers/reward_manager/naive.py": "e00bdfa1bf1418b302c79a4f8e5bc8e71a3350e797f525e524df5eff9fe3b61c",
+  "reward-signal-analysis/verl/workers/rollout/naive/__init__.py": "68176cf43c79bc96d5549d1950e402f391f39ddd62582ffaac62d649015af785",
+  "reward-signal-analysis/verl/workers/rollout/naive/naive_rollout.py": "f243226fbb0db0cf2d2e2c3c96730da81994ef358c2c06921c98564164356edc",
+  "reward-signal-analysis/verl/workers/rollout/vllm_rollout/vllm_rollout_spmd.py": "bc1685d0644f2a4168fb13438ad405cb3b8530fe839668a289d0264566da8970",
+  "reward-signal-analysis/verl/workers/rollout/vllm_rollout/vllm_async_server.py": "8235f369ec5015878626a46a490883f581d468977bdfdbfe6ee34d46aecc1d4f",
+  "reward-signal-analysis/verl/workers/rollout/vllm_rollout/__init__.py": "e64905d6e44d48462e57910c539fd9dd072519090dbfc2cd360732f663618225",
+  "reward-signal-analysis/verl/workers/rollout/sglang_rollout/utils.py": "1f42c0863f66c428c42850fd961ec3e5b1e1e150b58b7ade515ab9f5e1488e18",
+  "reward-signal-analysis/verl/workers/rollout/sglang_rollout/sglang_rollout.py": "dfa9aff827792281ed05f19ede80ce997c1f6445864a31c0adbb528766e34f44",
+  "reward-signal-analysis/verl/workers/rollout/sglang_rollout/__init__.py": "75183db302121275b3f66320e61ef9c82431c11b8c4945939fd64ea4f1f3455e",
+  "reward-signal-analysis/verl/workers/rollout/sglang_rollout/async_sglang_server.py": "6dd07bcea61321c7916c3051a6aaa23c5ccb9e4862d36134ed90f51a6e5a39e7",
+  "reward-signal-analysis/verl/workers/reward_model/megatron/reward_model.py": "aeacc2c5c3ee14a8aad46bace928234291fd9ef6d7239217e051055c60a05ac5",
+  "reward-signal-analysis/verl/workers/reward_model/megatron/__init__.py": "b820fe8d2c91d40953937619463b5bce5efc8e296725a673bfcfe41d5a262d6d",
+  "reward-signal-analysis/verl/single_controller/base/decorator.py": "2e4ffc5445c4969481be1750fb92dc1d40e79d93d8c1a56b5c98d0b86d821d75",
+  "reward-signal-analysis/verl/single_controller/base/worker.py": "5559485952d5078be15fcef9189600f668362ccbbf95b1a3ac47ba9daffc3bf1",
+  "reward-signal-analysis/verl/single_controller/base/worker_group.py": "93b26105ddc8ff6f5e2bfcb046cf42a11e01adf49749fc36a05fe011a813488d",
+  "reward-signal-analysis/verl/single_controller/base/__init__.py": "90b679f1bb261ca69ccc0ec2bec1fdae7372c3403521d82a420c3a6634e9b31b",
+  "reward-signal-analysis/verl/single_controller/ray/megatron.py": "c9c19a11b9cb191c5753fe844f7164a8e3db039fb92cb4b886ac7f5ce3fc7c10",
+  "reward-signal-analysis/verl/single_controller/ray/__init__.py": "0845f10bbf2ed64a0777aaa5b5eba575329befa88adba20559a5d2ce3821b2c9",
+  "reward-signal-analysis/verl/single_controller/ray/base.py": "c93f56df0daf5a703cb762d0da5d58193b1554f8b35164ed28718f2292a29501",
+  "reward-signal-analysis/verl/single_controller/base/megatron/worker.py": "ea1ea7c9a4ade34e5f0d2a4f6ac440115ebd3fa0e2d23359205b962b68f16ae4",
+  "reward-signal-analysis/verl/single_controller/base/megatron/worker_group.py": "2163689e2c403a821c444c6635bd1796ff3fd062a979a2cc7075c7a0cc1608e0",
+  "reward-signal-analysis/verl/single_controller/base/megatron/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/single_controller/base/register_center/ray.py": "94b1003e314155fd81905f91a3ab2fa8e259244edcec8d227b3f171db8035886",
+  "reward-signal-analysis/verl/single_controller/base/register_center/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/experimental/agent_loop/agent_loop.py": "04dde462c70b8a985e7a5716040588ec49dce4d4a5f40c39a18de4ef27b87c31",
+  "reward-signal-analysis/verl/experimental/agent_loop/tool_agent_loop.py": "fa02d61303fd066d769c03f476b89e43daf832e559f4dc981707b25a4377ce9f",
+  "reward-signal-analysis/verl/experimental/agent_loop/single_turn_agent_loop.py": "b7bd874682f9378fd4785d92c30d3d6b609c0d3eafca1035924dfe5c2c8d0d6b",
+  "reward-signal-analysis/verl/experimental/agent_loop/__init__.py": "27255930b36a316cfc6913e935f7c6e5e70a7643f84314bc99909e9e59170030",
+  "reward-signal-analysis/verl/trainer/ppo/ray_trainer.py": "df3ca89e4634009fbf53f5279b2f2dfa475cf9a88b6a80db162c342540a10feb",
+  "reward-signal-analysis/verl/trainer/ppo/ttrl_utils.py": "7935664548876b37c10b7216232271fa339cf90ef64f1aff74ca28d98cbef8e7",
+  "reward-signal-analysis/verl/trainer/ppo/metric_utils.py": "cd4ed003655fc58f9808def7723d23c49c79cc890cf7cc43a692aa04290bc151",
+  "reward-signal-analysis/verl/trainer/ppo/reward.py": "b887c287939511b38a1af07a0293ca3ca056387a7028379cdc61cd5b77532e71",
+  "reward-signal-analysis/verl/trainer/ppo/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/trainer/ppo/core_algos.py": "cb7365e6027e62f49814115ad973f0fd5dd7c07a43bf952f8797917bbf7247c3",
+  "reward-signal-analysis/verl/trainer/config/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/interactions/utils/interaction_registry.py": "653d7d4b7fe1c3d1fa9a26a0e0b0324252596ba09ebfe2aa658327e05b4ac5eb",
+  "reward-signal-analysis/verl/interactions/utils/__init__.py": "0bae18fab65cb4028732b91a1215baf8e02f9fdf8b4762a523662a34374e7432",
+  "reward-signal-analysis/verl/utils/kernel/linear_cross_entropy.py": "72a64da15663ff2b400fb7093e8388a5ecdebc14bce6b2c8c894696f15859405",
+  "reward-signal-analysis/verl/utils/kernel/kernels.py": "f9a7b49a1c617ccb06d436204b940bbf62accbe867be6cba5e5e9488c22d76a3",
+  "reward-signal-analysis/verl/utils/kernel/__init__.py": "db5168f476482dee94850c08859bc403f08f8eb0673d32a918b89b3ce9a430ef",
+  "reward-signal-analysis/verl/utils/megatron/pipeline_parallel.py": "b1bcc5ed48abd97f520b95f2c650d93e3a5f2c0c3518eb7baa4984c069f74781",
+  "reward-signal-analysis/verl/utils/megatron/memory.py": "b5e9dd018ea85a9d99ac80bc6bf7820b3d75cecd61fab4cfdfb9f78f5a992fdd",
+  "reward-signal-analysis/verl/utils/megatron/sequence_parallel.py": "8acb3406dd403eeab34a972e172982001c451e5e804ba0cf833a511a43f2a915",
+  "reward-signal-analysis/verl/utils/megatron/dist_checkpointing.py": "ef32128cc423fffaed869c71bcb85c0bba3a89b75648f4e1d311e585c041d34e",
+  "reward-signal-analysis/verl/utils/megatron/tensor_parallel.py": "53e71cea2d974bd1f68b9ca49c67dcc4e794601ad533ee15e11b02e8df222900",
+  "reward-signal-analysis/verl/utils/megatron/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/utils/megatron/optimizer.py": "2e477bee15a6c5a1c26c28fa026df928982038b13df9e8996916698ad3b53d1f",
+  "reward-signal-analysis/verl/utils/reward_score/math_verify.py": "0bf89ab36401642a42bfacc53b8a954d87b168a11fca0dc9014c9d1e21ecded3",
+  "reward-signal-analysis/verl/utils/reward_score/geo3k.py": "7e2821b85e2e07bf03e36f8fcd5f27faf2506e6f287098154b518c0a128b1e6c",
+  "reward-signal-analysis/verl/utils/reward_score/reasoning_gym.py": "840fff557b5e959054ab6c6e35f2936d03b645bdf80ce66f8efd23e9b1dd7003",
+  "reward-signal-analysis/verl/utils/reward_score/math_dapo.py": "4d8db99737fc392285772f4dd71dc54995ec57013f320a31d7ab027902e59d6f",
+  "reward-signal-analysis/verl/utils/reward_score/math_batch.py": "914e0e488bc876ea5926df958402fbf74eeee015356a3ddba3d0bdebdef24c85",
+  "reward-signal-analysis/verl/utils/reward_score/search_r1_like_qa_em.py": "8070e4d07719f23e626968b983c597918c73d15c6ab5b09c5346e7311d0cac83",
+  "reward-signal-analysis/verl/utils/reward_score/gsm8k.py": "5f1b6aba82138c0ea4e29ea430cd754480fb5372c807441f5043e6bbdb30e209",
+  "reward-signal-analysis/verl/utils/reward_score/__init__.py": "7c3b3634a99892297c0a0656e32fffd3b0ec843c1f06e1888b62999b6a40381e",
+  "reward-signal-analysis/verl/utils/reward_score/math.py": "2d48827903f42fc4bc0dbabd4652a5501d9a9b5bd3c4e6f32f06dc2fcd6993aa",
+  "reward-signal-analysis/verl/utils/checkpoint/fsdp_checkpoint_manager.py": "5ab42ccc8be63da5ec369e5c90489e5ed4251d232a5cf3179c6f370e13753608",
+  "reward-signal-analysis/verl/utils/checkpoint/checkpoint_manager.py": "94e1b1ca2f894df5a53dbe3b35adce07820a437dcda5b70a6bd2a6075405d7f4",
+  "reward-signal-analysis/verl/utils/checkpoint/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/utils/checkpoint/megatron_checkpoint_manager.py": "3e90e2c51dd2488baec759f9e949223f7450659442f17dfdc2d8a864ae6a05a5",
+  "reward-signal-analysis/verl/utils/rendezvous/ray_backend.py": "ae45c1c6e1a204b8c4d99849b03f1a571ce05a0e143a3b1dcc27522c7aee0ff9",
+  "reward-signal-analysis/verl/utils/rendezvous/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/utils/dataset/rm_dataset.py": "92b2913ef3d72811e209830d8aaba8903935b05aade88ccd10b7c8580d03278d",
+  "reward-signal-analysis/verl/utils/dataset/sft_dataset.py": "c4ffcb3e968141e72b78f76671dbb3006b6280dcb94b55349cc900cd235aaed5",
+  "reward-signal-analysis/verl/utils/dataset/epoch_sampler.py": "71fe83bf211eff335328cebe6ed5203cbbf563d475f3d887fe24d87d27e61858",
+  "reward-signal-analysis/verl/utils/dataset/rl_dataset.py": "dca6dba8253cb1bbfd6e037a6a449eace18a62cf15c2c1789f17fd21b31bf1ff",
+  "reward-signal-analysis/verl/utils/dataset/vision_utils.py": "122ec91d7944ddf4441f9db1d64043ba6b50ec82c6b600766e2b59d786b64d9e",
+  "reward-signal-analysis/verl/utils/dataset/__init__.py": "dd228db2fbd127fbad9d44ed15c4602a6979dcd67c33507c6c3f21c2b298beb1",
+  "reward-signal-analysis/verl/utils/dataset/multiturn_sft_dataset.py": "056234d2ce8fc2b8b14b73bc6f9813666462a1cc2e7ce8bf49837ef2bac232d2",
+  "reward-signal-analysis/verl/utils/experimental/torch_functional.py": "db137657bf06432600a9a2c570eff605e3b91ffb02035e9f13b76df8a2f60883",
+  "reward-signal-analysis/verl/utils/experimental/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/utils/metric/utils.py": "8937ff27a71ae8d1d1078c0353ae0e1269c514106f2507a34b03ad725ede16e6",
+  "reward-signal-analysis/verl/utils/metric/__init__.py": "0446c5125e52e296b4f3e54f766f0d0b9c5544648f02e6fecc1d83ea450be5e6",
+  "reward-signal-analysis/verl/utils/debug/profile.py": "b45be9ca3ce44611559fa90828555ef0f6c27f8df4c2d486a04842a97d679a5a",
+  "reward-signal-analysis/verl/utils/debug/trajectory_tracker.py": "2600fcdfca5a59fdc1a7ab63e41b8278d89ff47459bad1a78fdcebc348b3a72f",
+  "reward-signal-analysis/verl/utils/debug/performance.py": "4f9e840d67eb9b41d89d83577b34ea928294dbb3db5be1efe61bc843fdead76e",
+  "reward-signal-analysis/verl/utils/debug/__init__.py": "a2e7b5bc4c9e134b6cf019c0d4f204f8624c75de5cdc206c11c9a7be1b87d910",
+  "reward-signal-analysis/verl/utils/debug/empty_annotations.py": "19d56f53436bdd181b79a5dbee0dca77a232131a382883c302de30eaa2a80c30",
+  "reward-signal-analysis/verl/utils/debug/nvtx_profile.py": "2beceeebf2e45ac2871e48e0932cc128b42a1849c86776523b3e4dd3a3f15e93",
+  "reward-signal-analysis/verl/utils/logger/aggregate_logger.py": "50c9d25da02748122aa96521ca84dc83c934ddafa1480183a02c42ae53cab7e4",
+  "reward-signal-analysis/verl/utils/logger/__init__.py": "d4e7601565ea3c5126abd8e8b60751f62fdd3bdabf55f80eb9589abe7b342867",
+  "reward-signal-analysis/verl/utils/reward_score/ttrl_math/math_normalize.py": "048ed5e48921fcf40575bffad8701e6ac780b93ee3fabdd8846c761fd7021d07",
+  "reward-signal-analysis/verl/utils/reward_score/ttrl_math/math_utils.py": "417e696e9335613bc6a28860c0e86585656cd8a5fd00fcfb48edd95b799dc8cc",
+  "reward-signal-analysis/verl/utils/reward_score/ttrl_math/__init__.py": "6e7f8ea703258c051e4c28379443416a485046c235196f4ee25a244c216e994c",
+  "reward-signal-analysis/verl/utils/reward_score/ttrl_math/grader.py": "092573767fe18a60569a1ebedfc7c4e726fee0d23f1639f8de6dd4fa6f32fa77",
+  "reward-signal-analysis/verl/utils/reward_score/ttrl_math/cluster.py": "d343877bc4b7231788868246db97a55976b55f29650efa8e07688ac1d40c9c66",
+  "reward-signal-analysis/verl/utils/reward_score/prime_code/utils.py": "a85e7d0cb7f5caf2615c6d0861e695102f3c0e0a19ddae29efb1e726297aa67c",
+  "reward-signal-analysis/verl/utils/reward_score/prime_code/testing_util.py": "8e41c1d96c00aa1839cf3fdf59cdfcabf35c2f1e21a1206e5b0d014213308aba",
+  "reward-signal-analysis/verl/utils/reward_score/prime_code/__init__.py": "139690b83b2b21c0797696e07a36f24a206808183793d29fb15a5d49b8e58efc",
+  "reward-signal-analysis/verl/utils/reward_score/prime_math/math_normalize.py": "048ed5e48921fcf40575bffad8701e6ac780b93ee3fabdd8846c761fd7021d07",
+  "reward-signal-analysis/verl/utils/reward_score/prime_math/__init__.py": "bfaab493bcdd3d5eaf431e050b58339d63db5a71dae8e940bc58f232d2a9e407",
+  "reward-signal-analysis/verl/utils/reward_score/prime_math/grader.py": "092573767fe18a60569a1ebedfc7c4e726fee0d23f1639f8de6dd4fa6f32fa77",
+  "reward-signal-analysis/verl/utils/reward_score/sandbox_fusion/utils.py": "f9fdfafa5c23b276be6a836ac125191b339732ab995a80fbb52547d96952f3d7",
+  "reward-signal-analysis/verl/utils/reward_score/sandbox_fusion/__init__.py": "c2170fe3644c083b39d824b7fcd52149badf015a093a1389cded88e1258aeb4d",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/unit_test.py": "d6b700bd4ebd8304420ab8afba303f09ff6780cd6bcc0ac8c8e09fb4073a668f",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/compute_score.py": "7a680b2ed92825d3b08c03fd620a0ee05c388cc2e6f75f03a90da23f9ae45564",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/__init__.py": "1de19f5a6a2cad3656346cbee6375c199ef5feafc4e45390e8305ba5ad669649",
+  "reward-signal-analysis/verl/utils/reward_score/deepscaler_math/reward_types.py": "55d14b4d26080a9c508c2492713e4c4375a30b7c4d696d52da84b9e3d215e155",
+  "reward-signal-analysis/verl/utils/reward_score/deepscaler_math/math_reward.py": "e9b124732b63d0cdb1aa2578bbe77a20cde7f49c84fb16b915905f42b35de22c",
+  "reward-signal-analysis/verl/utils/reward_score/deepscaler_math/__init__.py": "ced614c4632940ecad45b76b8fa40e1f0645026caf1ab887d3005a7fea8c02b6",
+  "reward-signal-analysis/verl/utils/reward_score/deepscaler_math/globals.py": "4c53eabe983e2d4e477a6789419584c2a1fc7ec0c20591e4488f7ab37323b71f",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/lm_styles.py": "da5380d5544adf2aa718b511ab87499c876d078647b1a4f651ef19687900b677",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/benchmarks/test_output_prediction.py": "2493f00ae1841d0997f5a35632c3d447a189c68641d636bdaa4e91d16bd30d05",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/benchmarks/code_execution.py": "b891e798b34b40c32fcfbaedf683c714937f5e10014864e5fd7f3b9c9f6b077f",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/benchmarks/__init__.py": "f2ef3a9e8f0fca97923f1f1e937796630c766c69fee993a09289dd2f57eea0b9",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/benchmarks/code_generation.py": "0bc93f5a650b55d6a1b54d3db852d80ae64891ece4ff085b2e6b469eb89f8846",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/prompts/self_repair.py": "f337011aee38ecf69e5624ae362e48a1b40a253f0c40dd940c46cf85cde3eff7",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/prompts/test_output_prediction.py": "a1b32692f3fe1868821112591a3f06f01f2bac9124f2770f42b5720ad36fd71c",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/prompts/code_execution.py": "01ec4fd9703b79999398200f6804eac2147fe37e747157c025f49762899da6a8",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/prompts/__init__.py": "2c510da99fbfaf037d64194fbc8b2ce87c60c906d9f6a927148f18851bfbce0d",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/prompts/code_generation.py": "493ad74170f7e7f0f01490333524153524b024f10e9d9a28402cd4133445633f",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/scenario_router.py": "bc57cf68e3d4653ca62e964c78366571729d56768aac6fbfb56a6581b4ef9c50",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/fireworks_runner.py": "0ce25060a447dc289ebe3c1f5ee26a2c78d253cb8183d780ad18829e68d4fb5a",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/parser.py": "20cdb815b7390d93a4a4406338873525287030b9398aaa46188ca743af51d916",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/custom_evaluator.py": "458ed83c75e1fcc5b483670a208dbd655e84fb90e9a0516fa43c8ce28666f213",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/claude_runner.py": "1922b9dce7b042703ca46b6a283ef5fcc65a6875a17bda333500040d00f40359",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/deepseek_runner.py": "cc168d37744fe7d4e3d13f1e3e4e13ab834865906a3196e2cc20b3b6b8cb6064",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/base_runner.py": "68b9daffc20166e1a25fedd6ab1ee01347c15eacb18702bf51078d2fe2593282",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/cohere_runner.py": "8d8d528cf4fef0610583c70dc46b8f4ea469252660e3e4846d00db2dc93c3a4e",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/mistral_runner.py": "1dc819e27cdb62c7170f46c540db68e4f91a3791518158f59af45647dde8d14b",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/gemini_runner.py": "55dcb301f09db277a7a6d5da4db84dcb5ec3daa39ac4a0fa599c0d728fc03180",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/oai_runner.py": "fffb6b19f66ffb34f6f5eac61f7757365f9762d789c55e652151ef599579d8e6",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/main.py": "1ea338cccc2df08bfa0fd916d8e57da515f1be5458bccdfae98f422583591735",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/runner_utils.py": "577bf39742c7ecc154e07e44655443579c16f29acaa0eb8db0eb8ff1d691573f",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/vllm_runner.py": "4bf3de111cd7a457810f15c39fc292abcc9afab3ea1c7045e289235787131720",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/runner/claude3_runner.py": "f46ba4fef6886e240c143dc473e77aba684da3475b2bd10bfcb1c50a3d7bdb33",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/old_results_check.py": "346fed0487c18872b466dfc5208d1bc7de847689171c69e38e8e48189de76bd8",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/pass_k_utils.py": "6af3ee55d40f96fa06fa511bd5e43d94b19604750e82454b4b3c1f45c5199010",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/compute_scores.py": "c34d67dd5b3b298244c961ad8785e80e1bbb64a66aaffab24180d30f3d5260ec",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/compute_code_execution_metrics.py": "22f77cdd66f088e34403d64326970f977a1957505b958ee2327aef8c60d8ba77",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/compute_code_generation_metrics.py": "ad6137754bb31de164e41481bcb52caf06bb904ef2849245c34c83fc7ba7f45d",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/utils_execute.py": "fb86e4c0834ccedd3f057006410909aeb79d9970db54838a9bed79d7039624fa",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/compute_test_output_prediction_metrics.py": "8f6e61aa4cf94687e4f1756c04400df8093669b370390df2b9e1e1f5aba1aa55",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/testing_util.py": "cd06a6add04f144056b668ead55a0144aa6f94d7bc1f42d4b567e1a3c9d3e6be",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/evaluation/__init__.py": "673f2ba710506f31336369b641fc2f264a5216e962e3c6e81411f95bce879ca2",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/utils/scenarios.py": "3ff5d147fa6f56fde85e881a5ca359f502f983e8023f9a42cf87ff5043e263fc",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/utils/multiprocess.py": "13ad73a5b547acb41fc47192d34e836fc83e45abec6dde66b7744142638026b8",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/utils/path_utils.py": "d9edb9064af46a8234c610c42bb03fb91ba3c3a0ae8730f00ccb3809d5042360",
+  "reward-signal-analysis/verl/utils/reward_score/livecodebench/lcb_runner/utils/extraction_utils.py": "9e79c57cf3f0c6c25c3751adbb447979da9f6a614f46f52a62189722caea1f69",
+  "reward-signal-analysis/verl/utils/reward_score/deepscaler_math/utils/utils.py": "9bdc85ecb9c9255c387c04df5ae2d9a858a44d169418f05c262b78d2c33c963b",
+  "reward-signal-analysis/verl/utils/reward_score/deepscaler_math/utils/__init__.py": "095f4683aab6c0a465172625178dff889ff5bd15c29dfe0be00b57c25496464b",
+  "reward-signal-analysis/verl/models/qwen2/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/models/llama/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/models/transformers/llama.py": "f64a63a7e8f5b3c96d569acc5518c8296160e2e2a9ab1422436821ed9c9d98fa",
+  "reward-signal-analysis/verl/models/transformers/glm4v.py": "4d897e504910f204763b929b3509449c9c42fd737d696d2cf5664be2eadb911e",
+  "reward-signal-analysis/verl/models/transformers/qwen2_vl.py": "d16f9e29d3f3ffd4c8c0ab740a3a6e408524eb12580c5ef7c7b2cb88d67c449d",
+  "reward-signal-analysis/verl/models/transformers/qwen2.py": "e718af36c4b443e2bf4f98e4dcd4c61cc236caad709a3f92d170ad90194e495c",
+  "reward-signal-analysis/verl/models/transformers/qwen3_vl.py": "e02ed7305c42ea59c96189f4e8a350785b59a842a702ce43b46b69fb1095b815",
+  "reward-signal-analysis/verl/models/transformers/monkey_patch.py": "046288f97546fb0703e622618f3cc185d444867fab023ada52c5b694a6071473",
+  "reward-signal-analysis/verl/models/transformers/dense_common.py": "9f455a8af01290eae28f78ce5329175edf31c2425b02722567c350619c3b548a",
+  "reward-signal-analysis/verl/models/transformers/qwen2_5_vl.py": "fa7b4964a6b83c9eb2140166a15774ddd8fc1506cc12aa4e35b8ed2d76732c4d",
+  "reward-signal-analysis/verl/models/transformers/apertus.py": "f8af594815550ce344b390133da1eb97ee9b4a6225d3b08642e10b9156d6d946",
+  "reward-signal-analysis/verl/models/transformers/npu_patch.py": "a5a9d647d63a6e5f7a5f28f82c79af73edbeb1ea549ae3f889ecbd02cc6d7a90",
+  "reward-signal-analysis/verl/models/transformers/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/models/transformers/kimi_vl.py": "e38545c7093d68cf7100b63ddb9c70daf63bf5b1b61ba6ff1dd8fc330e393efc",
+  "reward-signal-analysis/verl/models/mcore/registry.py": "b02cf7b46c882cd31e447969ec6c94995e299f11c4a353d90ddf9d5caa2e9db8",
+  "reward-signal-analysis/verl/models/mcore/weight_converter.py": "2d6cf1ed29cf6cf0468dac4767fe450a98bba32495a35e80feb15713bdf403ce",
+  "reward-signal-analysis/verl/models/mcore/model_forward.py": "d7b2c98576beac26392ca0fecee8083c27d6c998d6b66f2cf071fdc7fcb57663",
+  "reward-signal-analysis/verl/models/mcore/patch_v012.py": "5e983d17aa0f67f0309ad26c6c9fe8c2bfef40691b91f162748454f2f26880ad",
+  "reward-signal-analysis/verl/models/mcore/model_forward_fused.py": "7213728449ec2c26b9ec67dba526bc2a175e0f3adce8f8ea53db8238ae023dc8",
+  "reward-signal-analysis/verl/models/mcore/config_converter.py": "07bc3ed39d99500fc6bff7851d009b775ee8f0426a446ef0c32eafd305edcaf8",
+  "reward-signal-analysis/verl/models/mcore/saver.py": "a82703fade74e67efe7e65dac451d25b0f1489ccd4dc689a618399ef051858b4",
+  "reward-signal-analysis/verl/models/mcore/model_initializer.py": "2a74e48dbdf528b59ebe29853f0de91f86aafbab8948b9dff3d623528c9060d9",
+  "reward-signal-analysis/verl/models/mcore/loader.py": "d07736ae2a75b5eb4944e5a28c3cc48a2f0459dd950903fb047c343225b8778e",
+  "reward-signal-analysis/verl/models/mcore/util.py": "3f717c3c6e20269e83aad80347da45803e760e179a06a37af31688804a474f0f",
+  "reward-signal-analysis/verl/models/mcore/__init__.py": "1138bedb8c7cc5741bcbe04bba765ed01871cbb311680ceb1e1da898e4f29a6b",
+  "reward-signal-analysis/verl/models/mcore/mbridge.py": "213ebe4dfa813b18bb48a18eb575669baa2df345c1428294bd32cd7208195954",
+  "reward-signal-analysis/verl/models/qwen2/megatron/modeling_qwen2_megatron.py": "7485fe813dc2eb4350681aafcade3097775f73674ae4f0cc8cff4ae9aef176f6",
+  "reward-signal-analysis/verl/models/qwen2/megatron/__init__.py": "8aee5796a576d05221db3a2fad83298ca997e9906a9d897575c1dc79455085c1",
+  "reward-signal-analysis/verl/models/qwen2/megatron/layers/parallel_rmsnorm.py": "a0c7dfaaa31622b2e5195e23c338ed466054cd22ca7e385bb19daae2ee426604",
+  "reward-signal-analysis/verl/models/qwen2/megatron/layers/parallel_attention.py": "0d329783d60f01e7839900554064910f63bd32caaab44a3231775ddc0fdca84d",
+  "reward-signal-analysis/verl/models/qwen2/megatron/layers/parallel_mlp.py": "2bd521a8cd29ce446e0322d469d0f74bf7f0903d983db3cf3a2c7cfa8d9a04c0",
+  "reward-signal-analysis/verl/models/qwen2/megatron/layers/__init__.py": "9f705565e0ce116d7bf2844f845c2e77ed8759d76d59b8b9b9ac94434d4398b7",
+  "reward-signal-analysis/verl/models/qwen2/megatron/layers/parallel_decoder.py": "be67a0f79d43f30fcb5a4878459bcc136cae71abee317967c6e8849416e9b082",
+  "reward-signal-analysis/verl/models/qwen2/megatron/layers/parallel_linear.py": "234bd3cdb743212308aeb8f501fc8b18e931d5c8b5b7367613bc24266ca08981",
+  "reward-signal-analysis/verl/models/qwen2/megatron/checkpoint_utils/qwen2_loader.py": "e9c79c8e11898f0eee4671ddd6bd8b2c669074078fd2987e4e9f3cd28f3da9d8",
+  "reward-signal-analysis/verl/models/qwen2/megatron/checkpoint_utils/qwen2_saver.py": "cc9e10740645e7e00c8d3a3f929fd9f7d2e23fb7747496b1e48a9755a906dea8",
+  "reward-signal-analysis/verl/models/qwen2/megatron/checkpoint_utils/qwen2_loader_depracated.py": "aa1fab9f73d274b1bc66dbcde02d96850872ba8c669b144cc58963322830bee6",
+  "reward-signal-analysis/verl/models/qwen2/megatron/checkpoint_utils/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/models/llama/megatron/__init__.py": "882e9d87f97eba0d0f67cacb2851e709c1771043e06c42880ccb11170d9914bd",
+  "reward-signal-analysis/verl/models/llama/megatron/modeling_llama_megatron.py": "0d7f09def7d9bcb833c09bb76038bae889f260c92a5e7cd66c15a371b095d361",
+  "reward-signal-analysis/verl/models/llama/megatron/layers/parallel_rmsnorm.py": "93f929f88639f8b8ab56f2a67e21dd94f4fac64ccead78de443c770728472ef5",
+  "reward-signal-analysis/verl/models/llama/megatron/layers/parallel_attention.py": "fafa53c0cbbfbabbb21e0cca09c9be45ce137f31800436685b0461c502a47cb8",
+  "reward-signal-analysis/verl/models/llama/megatron/layers/parallel_mlp.py": "d8cd610347584db7f6557c4324bce63a750af0a783168ba4d41e3f50af85af67",
+  "reward-signal-analysis/verl/models/llama/megatron/layers/__init__.py": "9cb3d2cecc91f210d111fd44f8122e652285265204832c22799428b41be8db9f",
+  "reward-signal-analysis/verl/models/llama/megatron/layers/parallel_decoder.py": "597288c158587c4a47f3e0e916bbeefb4e97d2085e850928ba482ad94f61a569",
+  "reward-signal-analysis/verl/models/llama/megatron/layers/parallel_linear.py": "cc74ba5c9c3dc6e7e7051dc3d6c1954dcdb204dd9a1e551e7cf91c017fa649d2",
+  "reward-signal-analysis/verl/models/llama/megatron/checkpoint_utils/llama_loader_depracated.py": "03c5c40fa17fc68b238951334e670ee592f5e356f7d35b7e0d353ea86170aac9",
+  "reward-signal-analysis/verl/models/llama/megatron/checkpoint_utils/llama_saver.py": "98ab4629ea317f20458ef5a850332ede5da4990416fac3da65adf012aa5ad3f7",
+  "reward-signal-analysis/verl/models/llama/megatron/checkpoint_utils/llama_loader.py": "842e23c86ebc1e327f230c6cb9e178d83df92a11900315ed23d71fbd099aa43a",
+  "reward-signal-analysis/verl/models/llama/megatron/checkpoint_utils/__init__.py": "e65abd77db034b4f29eb1268c24867f84919dbce22e28d1997e563ed1d94179c",
+  "reward-signal-analysis/verl/models/mcore/qwen2_5_vl/model.py": "bed786906b1049585ddfd0bc2794c745f52340903527b8f41911b6a91e227e41",
+  "reward-signal-analysis/verl/models/mcore/qwen2_5_vl/rope_utils.py": "9202aaa71b9167efde1fdc9a726015e05e44d217a2d62187325debd447a72fd2",
+  "reward-signal-analysis/verl/models/mcore/qwen2_5_vl/vision_transformer_block.py": "0041d05ef50a3301d3eeabfffaf5baf8d45abc3a83d54112724de53dfb5775d0",
+  "reward-signal-analysis/verl/models/mcore/qwen2_5_vl/vision_model.py": "19d68643babddbe29893d641783b490a807206fb7601121b25596b589ea97002",
+  "reward-signal-analysis/verl/models/mcore/qwen2_5_vl/__init__.py": "aef86e621b8b2e2c1c59334915e3e040b92e5e49df6574e5d20768d1bcf47828",
+  "reward-signal-analysis/verl/models/mcore/qwen2_5_vl/vision_config.py": "f173783cc6213543063e65035c5ca3c927d93d5a0899d5555fa2a3979d8f1a22",
+  "reward-signal-analysis/verl/models/mcore/qwen2_5_vl/attention.py": "37e7eafae287f59e1e83ab1b929fdc1657840f1601e6625af64e221bc8ebbfe0",
+  "reward-signal-analysis/examples/split_placement/split_monkey_patch.py": "77304f491261e2541f17b0a8bde70951a329ca36b540380034706e6af93020a9",
+  "reward-signal-analysis/examples/split_placement/main_ppo_split.py": "971e0ea87d55763608e1d18a398a1b02248df4dbbeab9ad0c237ae778b2fbf0d",
+  "reward-signal-analysis/examples/split_placement/run_deepseek7b_llm.sh": "d90fa046bb3b398a9828788b997738a17088b31bff8d0ba6afb18e7a2f958eeb",
+  "reward-signal-analysis/examples/data_preprocess/preprocess_search_r1_dataset.py": "7bceaaf07bf8a44f8d5af0b52420aef0117b13ebcc8ae49ca8992357e511645f",
+  "reward-signal-analysis/examples/data_preprocess/polaris_math.py": "ad030d8fd453b8ac802f350e9f4de9bfa3a910befbd6f244917510f807947edf",
+  "reward-signal-analysis/examples/data_preprocess/aime2024_multiturn_w_tool.py": "46907948076d8f72f0369f89545a7fc5fd3e573b5af67e12bc93a450446ff926",
+  "reward-signal-analysis/examples/data_preprocess/geo3k_multiturn_w_tool.py": "2e564f220801341f95f540ef84e48b648813cf5e888d2ff8188e161386896516",
+  "reward-signal-analysis/examples/data_preprocess/geo3k.py": "2c3fe29196512c21ace8906c486c7e5eadaaaba612e9c6606258630625249a0e",
+  "reward-signal-analysis/examples/data_preprocess/dapo_multiturn_w_tool.py": "5a8cd42980f3c7f81309d870acefa595be8be9e3c1c903c00323f7151c5a24da",
+  "reward-signal-analysis/examples/data_preprocess/hellaswag.py": "9e847801a808f5074a385e17030f2cc2f3994893a0845ec9cd1e354d81b8b2b9",
+  "reward-signal-analysis/examples/data_preprocess/full_hh_rlhf.py": "aa05fd5c793388cf07582e9fa7a7c2bac7d34b0adb608c2c6c33130b3182d681",
+  "reward-signal-analysis/examples/data_preprocess/polaris_math_full.py": "8e0432541dc789deb4dfe1e48bc608d42b213dbaa02c4bc2e5a7e35d9d347dbf",
+  "reward-signal-analysis/examples/data_preprocess/gsm8k_multiturn_w_interaction.py": "43d9e2738596406ae98f0ba33634e47a3ea331a89341f84297c7bcccc8acd562",
+  "reward-signal-analysis/examples/data_preprocess/gsm8k.py": "0c37550447e212ac017f24335943dbdc9327d8eab3cc40884c64f4187a40d1b0",
+  "reward-signal-analysis/examples/data_preprocess/gsm8k_multiturn_w_tool.py": "2b1fd24c4b26e40fc7ed14d9a38438024b546260bd1c5eace48a92134895b121",
+  "reward-signal-analysis/examples/data_preprocess/math_dataset.py": "593f64f488a78d31bfeffa5973b3d7a22a5977cfa45cc6d19e14c302f0b54ed1",
+  "reward-signal-analysis/examples/data_preprocess/make_polaris_eval_sets.py": "56511030c76b7b74dc377550631db060da1c638c224f6ed167e55016fb9ca686",
+  "reward-signal-analysis/examples/data_preprocess/multiturn.py": "6789ca1900ee6fa97107929498fbdf43e061d180c2a267933757d1be652f7698",
+  "reward-signal-analysis/examples/tuning/70b/qwen2-70b_grpo_32_h800_fsdp_vllm.sh": "199c21adc9715feebf9bf4482018ea4ffe117d1db9c12de1eca590285b7c8ad2",
+  "reward-signal-analysis/examples/tuning/70b/qwen2-72b_grpo-lora_8_h100_fsdp_vllm.sh": "3d3138419bc71b819f9ffb048dd261bfa47a172e6c23bc98c73336ae90ee5f30",
+  "reward-signal-analysis/examples/tuning/70b/qwen2-70b_grpo_32_h20_fsdp_vllm.sh": "5515c3244368a83aee9185b9b3f1389ad80801233a0d644318e6d486b86c6e0d",
+  "reward-signal-analysis/examples/tuning/32b/qwen2-32b_grpo-lora_4_h100_fsdp_vllm.sh": "6baeabdf992772f14ab86876d89727d9ea4f537bc697f9cb4b6978a43342fb73",
+  "reward-signal-analysis/examples/tuning/32b/qwen2_32B_grpo_8_h20_megatron_vllm.sh": "ad2859a6424dbd2aada6c6e94b039e9783c198335950c7e0a7cd84a2e98b7beb",
+  "reward-signal-analysis/examples/tuning/0.5b/qwen2-0.5b_grpo-lora_1_h100_fsdp_vllm.sh": "f63fe06f7b108f93a9bc73b7d52d36c3eb0347e39f39d67975a1057afb0e2106",
+  "reward-signal-analysis/examples/tuning/14b/qwen2_14b_grpo_4_h800_fsdp_vllm.sh": "cffc86b63eac99c372c02becd5dc52585a2df5331627a0014aa2363eadb738f8",
+  "reward-signal-analysis/examples/tuning/14b/qwen2-14b_grpo-lora_2_h100_fsdp_vllm.sh": "c6dfaaaf1c01a04fa50e8fbc299a84df2f3aefc2cca13cb42ca642d8a0b06eb1",
+  "reward-signal-analysis/examples/tuning/1.5b/qwen2-1.5b_grpo-lora_1_h100_fsdp_vllm.sh": "032ae20cc426b068db968bc3d307c01886fdc1e609de53cd097b97ac5dd12cba",
+  "reward-signal-analysis/examples/tuning/7b/qwen2-7b_grpo_2_h800_fsdp_vllm.sh": "bea6a4041fecf3259603e943d7839f214bb21d14215930e36ab5eeb0bb976424",
+  "reward-signal-analysis/examples/tuning/7b/qwen2-7b_grpo-lora_1_h100_fsdp_vllm.sh": "9361d0202bb1718adfff50f118c7cd974d615f49857f1e020b5d8c749c1af895",
+  "reward-signal-analysis/examples/tuning/3b/qwen2-3b_grpo-lora_1_h100_fsdp_vllm.sh": "abd62ebf3dc28c0e0313e63ba2646d1492ff7b8205dfbe65ade887942fbce344",
+  "reward-signal-analysis/script/slurm/gpu_heartbeat.py": "362550e9384594690f5fc95200d611c2e827b72e35456ea8b8fab4d96acd9612",
+  "reward-signal-analysis/script/rl_with_replay/code/eval_mbppplus.sh": "2f8840ee55a6ca548c99e490c09ec8216c1ac59bce06be453bce3660666018b1",
+  "reward-signal-analysis/script/rl_with_replay/code/plot_mbppplus_passk.py": "6d5fb161359e2e78513f280251ebaf40f547451fc83281c080e8c742367a5a99",
+  "reward-signal-analysis/script/rl_with_replay/code/plot_dynamics.py": "483b3032bb8d186231bb43ce37cbe1db3696709ae583e51da8554e316f643340",
+  "reward-signal-analysis/script/rl_with_replay/code/test_reward_e2e.py": "4e39fa0f0d15d8faecfa50a9ac911cd3d4bdca767261cfdbae857038df656183",
+  "reward-signal-analysis/script/rl_with_replay/code/learning_forgetting.py": "5a8d4dc2c7b924295df8b824ba5762e6ab811fcafa725bbca2577b1c0410d613",
+  "reward-signal-analysis/script/rl_with_replay/code/skywork_nobuf.sh": "45ffd2195b746e9045c4a8e90f77f4ece26790a3e032aa927a7584debaf2d536",
+  "reward-signal-analysis/script/rl_with_replay/code/requeue_failed_evals.sh": "abcf79674933ff6e54e755d4f6e0a3aab0a7c9b86c991d6f6224c31db97cbd8d",
+  "reward-signal-analysis/script/rl_with_replay/code/eval_mbppplus_gcr_sft_checkpoint.sh": "4ac9af5b6a13ac966453f69c83771f9eb1e5cdf3c585c43bedb0dd5b090fa64b",
+  "reward-signal-analysis/script/rl_with_replay/code/submit_eval_curves.sh": "84d589b400dff19333887f0ac96e78f2d6a5abb767ffe396091557ddb7eee11a",
+  "reward-signal-analysis/script/rl_with_replay/code/restore_amulet_runtime.sh": "ae6187ee32070811d6f27c2b22594e7f04c5facd32b9fd7958365d728409961c",
+  "reward-signal-analysis/script/rl_with_replay/code/mbpp_hardcd_lam0p1.sh": "399cbe62ec64cd65c657fca6cae49541742e29ddbd19efbeaf761d48237321d3",
+  "reward-signal-analysis/script/rl_with_replay/code/mbpp_nobuf.sh": "49e12ce5dcc9f738157f7f9342cc4da64a5b467c5dbb84ad7935784b40dc00db",
+  "reward-signal-analysis/script/rl_with_replay/code/eval_mbppplus_gcr_checkpoint.sh": "453bfe8247899a7dd48fefd44ba4295296e1ed067d7bcc634a1441cd92dfadf5",
+  "reward-signal-analysis/script/rl_with_replay/code/export_prompt_table.py": "a63de6135051e9f621a9f11749e73ede2564439abf72b90b703958995d937669",
+  "reward-signal-analysis/script/rl_with_replay/code/validate_mbppplus_sft_eval.py": "05ab820723e1953dca13e8359d2b4cc374cc69334b0767c8cef1209e11b477d0",
+  "reward-signal-analysis/script/rl_with_replay/code/prepare_skywork_code.py": "5fe6ffc74484d4925d3f0123c65a31e28aa4f03b30a2cf8d3439afc2552f1964",
+  "reward-signal-analysis/script/rl_with_replay/code/skywork_hardcd_lam0p1.sh": "b27fdcb43fbcb1d1a5c419e4e3db72d7760e57cc521aec8a7b22755709e5f956",
+  "reward-signal-analysis/script/rl_with_replay/code/merge_eval_shards.py": "cadab7c011749e4eadf60d0ce812942eb9f4777ab20979b6d16846ee34032641",
+  "reward-signal-analysis/script/rl_with_replay/code/shard_eval_set.py": "4bd5569dfe3af2e4addada198aee43048dbd722e65fb35aecc77bc9c42bde608",
+  "reward-signal-analysis/script/rl_with_replay/code/plot_story.py": "65d28eeae0d532f7c1498483c221a9eb4a1c51b5625e584233c7ec2c5246120b",
+  "reward-signal-analysis/script/rl_with_replay/code/prepare_mbpp.py": "c190788828e5e8e4a6b4b5dc19185981ea3cd55b737943ae9e3e32ac3dbdc3f5",
+  "reward-signal-analysis/script/rl_with_replay/code/eval_mbppplus_vllm_direct.py": "e8104318e2ef7cdb34b7c9a92ce7faf33d9eb70f04719f6c44ef54c91aefba1f",
+  "reward-signal-analysis/script/rl_with_replay/code/_code_train_base.sh": "e92a9e568d1d7f5fd6411c16ba68416af3c8a3960ac24cb49af7051f34b2938a",
+  "reward-signal-analysis/script/rl_with_replay/code/code_prompts.py": "df3ce00751fe9235b9fb6db120a5be6be5dc347dee3dcb6298a75af9d06b18c1",
+  "reward-signal-analysis/script/rl_with_replay/kk/qwen3b_train.sh": "4bd5124edf4d285e734e3705bf7ae3905c15a662c9e0b541e751b772865ad37c",
+  "reward-signal-analysis/script/rl_with_replay/kk/qwen3b_train_nonformat_buffer.sh": "19fc8e40b5382bd98e36acef65e2dcb8207018e8c143c16a8a074dfcf7491efa",
+  "reward-signal-analysis/script/rl_with_replay/math/qwen_3b_buffer.sh": "b4d3564110640c906cb9984d7e19c5e960032a91af2e123f740df963ab20ea81",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/qwen3_8b_nobuf.sh": "e3350a43bbcfde38df11cae609c4386fd04561afe66597b79ab765bf8188656f",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/eval_llama3p2_3b_nobuf.sh": "3c4ce6fc5f767c99056aaad1e3f471444a3335b7adb6204187186bfe6438bf28",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/common.sh": "67276cd35de15327973b18f44e14179113862f69c2453118db45da33d0e79112",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/merge_polaris_8gpu.py": "62a5dbb9483ff6f9ec17be3e44d0a6ae5b3b51b5e9eb2c0df3afdf506577033c",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/fetch_polaris.py": "9136ddcd0f9f4fa6135d1f20b8408cb0c99f54c070aea3bd6033845640c66509",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/llama3p2_3b_nobuf.sh": "83b558fd4a5fb354d56d7407002a533e87609703a42ac08107ac43d84a6c28cc",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/llama3p2_3b_lam0p1.sh": "0c9093f29d1c24aecee6bf6e79b7e555683dde0abf1280c7f9d2fdbab7ab90a8",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/eval_llama3p2_3b_lam0p1.sh": "89918a13ac8351d485c09db5b35822c41beea16de57ac9d7f2f51bf834382536",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/eval_common.sh": "fe1912d5852b0acf5b96663cf565e82c18a9b784933c3ba6d3e576211b9a6f5c",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/qwen3_8b_lam0p1.sh": "48133c8f834e791f5d3c494a2c6004521ecdab8dca738d641179962d5a63b4f2",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/eval_qwen3_8b_lam0p1.sh": "5faaf88da3a07b032861be054ee97c2ee78dbd052ede6d1d4511b9f9ba34c767",
+  "reward-signal-analysis/script/slurm/polaris_8gpu/eval_qwen3_8b_nobuf.sh": "16dddc14f0f961a911ce60abdbf239dfeb6d05f7b119df6511ba533dc7a9d948"
+}

evaluations/mbppplus_n160_t0.6/provenance/source-training-proof.json

@@ -0,0 +1,42 @@
+[
+  {
+    "status": "training_complete",
+    "arm": "llama3b_code_ordered_lr1e-5_bs256_seed42",
+    "source_job": "llama3b-code-ordered",
+    "azure_id": "sad_drawer_63s9dw41sd",
+    "global_step": 65,
+    "checkpoint_steps": [
+      6,
+      13,
+      20,
+      26,
+      32,
+      39,
+      46,
+      52,
+      58,
+      65
+    ],
+    "final_matches_step65": true
+  },
+  {
+    "status": "training_complete",
+    "arm": "llama3b_code_shuffled_lr1e-5_bs256_seed42",
+    "source_job": "llama3b-code-shuffled",
+    "azure_id": "willing_napa_b29yxdlm8n",
+    "global_step": 65,
+    "checkpoint_steps": [
+      6,
+      13,
+      20,
+      26,
+      32,
+      39,
+      46,
+      52,
+      58,
+      65
+    ],
+    "final_matches_step65": true
+  }
+]