{
  "schema_version": 1,
  "adapter_key": "progressive-contraction",
  "validated_at": "2026-10-14",
  "seeds": [
    42,
    43,
    44
  ],
  "accelerator": {
    "vendor": "NVIDIA",
    "model": "NVIDIA-A100-SXM4-80GB"
  },
  "command": [
    "python",
    "scripts/run_oct13_post_training.py",
    "--method",
    "progressive-contraction",
    "--config",
    "data/oct13_execution.json",
    "--seed",
    "42",
    "--output",
    "runs/complete-progressive-contraction-s42"
  ],
  "command_path_relocations_only": true,
  "dataset": {
    "verified": true,
    "test_read": false,
    "source_id": "openai/grade-school-math",
    "source_revision": "3101c7d5072418e28b9008a6636bde82a006892c",
    "source_sha256": "17f347dc51477c50d4efb83959dbb7c56297aba886e5544ee2aaed3024813465",
    "partition_sha256": {
      "train": "6b8672439c8cf6f6b6e65bf5f0e515fbf4c497c9066eefb62a6a6a61ced158e3",
      "validation": "2cf1119ad7b2e31e825274487ef32c073e03a8ed44f08ecd3e0366c1f40ebdb7"
    },
    "partition": "normalized-question sha256 unsigned-prefix / 2^64 < 0.1 validation"
  },
  "checkpoint": {
    "policy": {
      "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
      "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
      "verified_mirror": {
        "public_id": "Qwen/Qwen2.5-0.5B-Instruct",
        "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
        "file_sha256": {
          "config.json": "18e18afcaccafade98daf13a54092927904649e1dd4eba8299ab717d5d94ff45",
          "generation_config.json": "e558847a8b4402616f1273797b015104dc266fe4b520056fca88823ba8f8ebe6",
          "merges.txt": "599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
          "model.safetensors": "fdf756fa7fcbe7404d5c60e26bff1a0c8b8aa1f72ced49e7dd0210fe288fb7fe",
          "tokenizer.json": "c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
          "tokenizer_config.json": "5b5d4f65d0acd3b2d56a35b56d374a36cbc1c8fa5cf3b3febbbfabf22f359583",
          "vocab.json": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
        }
      }
    },
    "teacher": {
      "public_id": "Qwen/Qwen2.5-7B-Instruct",
      "revision": "bb46c15ee4bb56c5b63245ef50fd7637234d6f75",
      "verified_mirror": {
        "public_id": "Qwen/Qwen2.5-7B-Instruct",
        "revision": "bb46c15ee4bb56c5b63245ef50fd7637234d6f75",
        "file_sha256": {
          "config.json": "7463bb0ea78315365e6c6b74de4e73bbcc8359dfb0c5a737584e077d42c0b03c",
          "generation_config.json": "3a8f9087e486054c8a4a08dae2e5a3ba62e23da212b5b8c08bc42cb983c3459f",
          "merges.txt": "599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
          "model-00001-of-00004.safetensors": "a1333e6293854747c481288ea83b348226af178dd565c49b6f9495ba1966aba7",
          "model-00002-of-00004.safetensors": "f5d25a2772cb825164a2a2c0fb6d51a87e282abf21e4dd75bc5cfb3cd0ea6185",
          "model-00003-of-00004.safetensors": "8efdec4c1bc12317ae1a38dc42b595ce777738a64deea3fcb8a0a91381bcdfd5",
          "model-00004-of-00004.safetensors": "1a72d403cdf0c1ec3cb7f289f17b394a01e64394c2e9b3c0f94dbce3faf879bd",
          "model.safetensors.index.json": "624bf7c47cd12468fdc16e38a47cf4f19e0415b859a223ba3c027eed2f0e1028",
          "tokenizer.json": "c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
          "tokenizer_config.json": "5b5d4f65d0acd3b2d56a35b56d374a36cbc1c8fa5cf3b3febbbfabf22f359583",
          "vocab.json": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
        }
      }
    }
  },
  "result": "passed",
  "metrics": {
    "initial.accuracy": {
      "mean": 0.5,
      "std": 0.0,
      "n": 3
    },
    "method_validation.accuracy": {
      "mean": 0.5833333333333334,
      "std": 0.14433756729740643,
      "n": 3
    },
    "verified-full-sft.accuracy": {
      "mean": 0.5833333333333334,
      "std": 0.14433756729740643,
      "n": 3
    }
  },
  "execution_checks": {
    "per_seed": {
      "42": {
        "parameters_changed": true,
        "measured_cost": {
          "generation_calls": {
            "evaluation": 12,
            "teacher_backbone": 13,
            "teacher_profile": 8
          },
          "generated_tokens": {
            "evaluation": 4118,
            "teacher_backbone": 2948,
            "teacher_profile": 4761
          },
          "scored_response_tokens": {
            "student_sft": 3191,
            "student_routing": 776
          },
          "frozen_encoder_forward_calls": 0,
          "frozen_encoder_input_tokens_including_prompt": 0,
          "scope": "entire invocation, including training controls, profiling and evaluation",
          "includes_rejected_and_profiling_generations": true
        },
        "method_updates": 3,
        "control_updates": 3,
        "verified_profiled_corpus": 4,
        "teacher_uncertainty_scores": [
          0.0,
          0.0,
          0.0,
          0.0
        ],
        "teacher_pool_indices": [
          0,
          1
        ],
        "exposure": {
          "method": 4,
          "full_sft_control": 12,
          "relative_to_verified_full_sft": 0.3333333333333333,
          "teacher_profiling_cost_excluded_from_ratio_but_reported": true
        },
        "exposure_ratio_is_total_compute_saving": false,
        "invocation_elapsed_seconds": 217.54227962344885,
        "configured_wall_cap_seconds": 1800
      },
      "43": {
        "parameters_changed": true,
        "measured_cost": {
          "generation_calls": {
            "evaluation": 12,
            "teacher_backbone": 13,
            "teacher_profile": 6
          },
          "generated_tokens": {
            "evaluation": 4118,
            "teacher_backbone": 3131,
            "teacher_profile": 4293
          },
          "scored_response_tokens": {
            "student_sft": 2616,
            "student_routing": 406
          },
          "frozen_encoder_forward_calls": 0,
          "frozen_encoder_input_tokens_including_prompt": 0,
          "scope": "entire invocation, including training controls, profiling and evaluation",
          "includes_rejected_and_profiling_generations": true
        },
        "method_updates": 3,
        "control_updates": 3,
        "verified_profiled_corpus": 3,
        "teacher_uncertainty_scores": [
          0.0,
          0.0,
          0.0
        ],
        "teacher_pool_indices": [
          0
        ],
        "exposure": {
          "method": 3,
          "full_sft_control": 9,
          "relative_to_verified_full_sft": 0.3333333333333333,
          "teacher_profiling_cost_excluded_from_ratio_but_reported": true
        },
        "exposure_ratio_is_total_compute_saving": false,
        "invocation_elapsed_seconds": 214.301582518965,
        "configured_wall_cap_seconds": 1800
      },
      "44": {
        "parameters_changed": true,
        "measured_cost": {
          "generation_calls": {
            "evaluation": 12,
            "teacher_backbone": 13,
            "teacher_profile": 10
          },
          "generated_tokens": {
            "evaluation": 4228,
            "teacher_backbone": 3103,
            "teacher_profile": 5830
          },
          "scored_response_tokens": {
            "student_sft": 3652,
            "student_routing": 690
          },
          "frozen_encoder_forward_calls": 0,
          "frozen_encoder_input_tokens_including_prompt": 0,
          "scope": "entire invocation, including training controls, profiling and evaluation",
          "includes_rejected_and_profiling_generations": true
        },
        "method_updates": 3,
        "control_updates": 3,
        "verified_profiled_corpus": 5,
        "teacher_uncertainty_scores": [
          0.0,
          0.0,
          0.5,
          0.0,
          0.375
        ],
        "teacher_pool_indices": [
          2,
          4
        ],
        "exposure": {
          "method": 4,
          "full_sft_control": 15,
          "relative_to_verified_full_sft": 0.26666666666666666,
          "teacher_profiling_cost_excluded_from_ratio_but_reported": true
        },
        "exposure_ratio_is_total_compute_saving": false,
        "invocation_elapsed_seconds": 228.65010682865977,
        "configured_wall_cap_seconds": 1800
      }
    }
  },
  "limitations": [
    "Three actual full-definition seeds42/43/44 complete this reduced-scale default-seed plan, not the entire multi-method batch.",
    "Three seeds remain L1 public-data mechanism diagnostics; no formal capability, stable relative gain or original benchmark claim.",
    "Tiny validation and realized costs are retained, including negative metrics; no test-selected adjustment.",
    "Only four validation cases; official test content was not opened or used for selection.",
    "Actual token cost is retained; matched configured caps do not imply identical realized compute.",
    "Qwen2.5 and tiny GSM8K train-only public holdout replace the original scale and Qwen3.5 models.",
    "Actual MCRT sampled continuation costs are reported separately; exposure reduction is not total compute saving.",
    "Published v1 HTML has missing appendix labels and no retrievable appendix algorithm; main Eqs1-8 are implemented."
  ],
  "executed_source_sha256": {
    "scripts/prepare_soar_catalog_qa.py": "483a4eb3d9932d31046888c165c06590e67c6df5575a5b2afd4e302145f445e7",
    "scripts/run_oct13_post_training.py": "916b3b9a0eaeeaf30bad811376fa05305011754c922bb368a0f39833aad07ffa",
    "src/auto_research/oct12_public_sources.py": "4e1150cbe4ea975819c244267180f853c83c4ad4c2c27f9e0ce95cc16793e093",
    "src/auto_research/post_training/oct10_checkpoint.py": "247f1f97a3ca701277f241ec81d99296f0ea82e1db8f68006fcd7bfdc661cd74",
    "src/auto_research/post_training/oct13_common.py": "739f276633d74a85ba790e4d6df25504bce89950c40867ed57302cde03f1ada9",
    "src/auto_research/post_training/proximal_checkpoint.py": "5307a54204720e5407d2cd0435b654ae0e594f30c85cdf07d55d4e90612a1397",
    "src/auto_research/post_training/response_policy.py": "e380a5b4d1d53974b973b72b7d9ad6328129dc8c9bae4d233f6c1cbdf8508bf7",
    "src/auto_research/public_holdout.py": "033d9c1fd112accf5ac79573b99d9f31cffc646bcab6911cf3e49459fa003b46",
    "src/auto_research/reproductions/gppo_framework/__init__.py": "7fe8cab42fef9f32b9653dfb5f54f03182de0753de2c9e5b359d124d7cd763d8",
    "src/auto_research/reproductions/gppo_framework/adapter.py": "a7f4ee792a8ded3daeeb183cbca5c99de63c1b7d3cb195f5bedd5289cb0e0260",
    "src/auto_research/reproductions/gppo_framework/objective.py": "0f8fbb2ad397b04a70466dde4e939f4d094ad7e7df07cfbba59f4a6ca5d05dd0",
    "src/auto_research/reproductions/gppo_framework/paper.yaml": "a023ee98bbd078077b2708122973d27a98085848be816ce461b7d4291aca8e1f",
    "src/auto_research/reproductions/grpo_difficulty_curriculum/__init__.py": "d09d8d7e0e6f5bcec9979241eed06a1d2a6ac8f3fb7a25f49e0df55be0071acb",
    "src/auto_research/reproductions/grpo_difficulty_curriculum/adapter.py": "bbd3e964dc035ea4bbd999a581ea07eccabd986b6456774b21a4c24092cbdfbe",
    "src/auto_research/reproductions/grpo_difficulty_curriculum/algorithm.py": "4131e39279d30c0748b14f611ab657b3fea76147a7c164aea2883c62bde75785",
    "src/auto_research/reproductions/grpo_difficulty_curriculum/checkpoint.py": "d03a7d2db0a19b469fe5429ddce634b02d8cbd0eeed8d09bef0afc51c1d7ac70",
    "src/auto_research/reproductions/grpo_difficulty_curriculum/paper.yaml": "0b4dd86c85907b732df9380a4bd1bbc17785c09c184d5be2793ac37f63768c16",
    "src/auto_research/reproductions/magicbeanstalk/__init__.py": "8dd207ea37a89dfa34f87b02eb4a2fd26472895678d1362a7e5b59dad5cda616",
    "src/auto_research/reproductions/magicbeanstalk/adapter.py": "319b6503de342feee49fb6446b7bbdeb117f7dc0dd7b79ed1c3bd34cce1ce73f",
    "src/auto_research/reproductions/magicbeanstalk/algorithm.py": "78a68baad5b98b402fb87702ab50d5e01000f41250111604b84321e6c3e4900f",
    "src/auto_research/reproductions/magicbeanstalk/checkpoint.py": "24dd8eb1979c04c2128919533f83fb347d9b40ab52184f7b1b9b3644512e9eae",
    "src/auto_research/reproductions/magicbeanstalk/paper.yaml": "e9c5619918b4d72d776008aba3fa95d50ba6ebe99afab729ba9e6dfe82fac06e",
    "src/auto_research/reproductions/progressive_contraction/__init__.py": "e50fe46d888e0a0b11b7d4d6d8f3224ae5940394f8ba8396fc5afb1cd38a4d53",
    "src/auto_research/reproductions/progressive_contraction/adapter.py": "06f23b5c795f7222f76dee9beaa8164f460eb5183e18c6d374bff011737235dd",
    "src/auto_research/reproductions/progressive_contraction/algorithm.py": "ba077b8c52a8d10e268a4e5c85770b9cebcb12415f9b2aaf7b9500e1303ca4b1",
    "src/auto_research/reproductions/progressive_contraction/checkpoint.py": "67fa44814d613d3838716fea68476217d608af14f23a051e32a916b72538e7ec",
    "src/auto_research/reproductions/progressive_contraction/paper.yaml": "1beccb702ce7a33b196b1fcc6bab13cdc32fc48952d247565d54e8ce830860c2",
    "src/auto_research/reproductions/soar/__init__.py": "3d87d1189d7037e32c1e2cb0c249d13609c5af3d7062a19904590ba7923c799a",
    "src/auto_research/reproductions/soar/adapter.py": "8fbbe8f8fe54b9f014ccc26a75b01e72b73b95e2b03cc379c5b5db4b63c9b31e",
    "src/auto_research/reproductions/soar/algorithm.py": "9b7068d9774d1c93800c512a04b9ba4d0ee98fc13d308e5ac0970299c8bb2485",
    "src/auto_research/reproductions/soar/checkpoint.py": "43a83930af442dc3610d61becd06b38ca3c921131ed417837d3d7e11aa672708",
    "src/auto_research/reproductions/soar/paper.yaml": "2f756cab9f7fefd4613bb69971fd8252c2b9ad451fa5c5eff5d03fa063d91521",
    "src/auto_research/reproductions/state_opd/__init__.py": "8db07736f0336ed7f39502ae2953bdae76658942542cacbe4b6b2c1fd0abe094",
    "src/auto_research/reproductions/state_opd/adapter.py": "2f8bf4773e8f40c0eabe777fbd1e1aa6d372a7b05501404c3e3f7b2d54987524",
    "src/auto_research/reproductions/state_opd/algorithm.py": "dc7fe3eae619ff69ec51c47826f89471f00be564ea4e6730e12726a060ba7e2f",
    "src/auto_research/reproductions/state_opd/checkpoint.py": "95755e0604f563a5c96f92e9de9069befe3e1274d31c224a43ca20d71f756b09",
    "src/auto_research/reproductions/state_opd/environment.py": "0e9564f9ec540985eb010b1e70f46938acc1bd946041b518d6579253762080e1",
    "src/auto_research/reproductions/state_opd/paper.yaml": "86c75255f1c416611a181c26359e9f5d9865caaea5266ca7b13dd8a183b906cc",
    "src/auto_research/reproductions/synthself/__init__.py": "474bb75e10b95afe6faec1f3631beab2f17f9a43013d3d29e79b7d4b36e06031",
    "src/auto_research/reproductions/synthself/adapter.py": "db4c08e1ccc2acfc28d19fc72acfa39b0a0dce096bbc430bafe9b21dbb72cddd",
    "src/auto_research/reproductions/synthself/algorithm.py": "1e7ea1ff65d3ba0ece1af889022c19f7147f51d10fb8196ab8f9fa2e18604b95",
    "src/auto_research/reproductions/synthself/checkpoint.py": "dbcf9f222094c197e0392f82191a0ae6551c6573ee54a9eee8c7d37553472981",
    "src/auto_research/reproductions/synthself/paper.yaml": "464d06e9cdd535ab61743fdba6641ae2eac4c8345fad27748c63e620544074c7",
    "src/auto_research/reproductions/tapo_trajectory/__init__.py": "a0dd8d4825b91dd076b71bdd3a3eccb0638ba4bf0240fe9da76d61a7290f69c6",
    "src/auto_research/reproductions/tapo_trajectory/adapter.py": "c01b75bf7853546caa2d517dd5f9c9faf739537b95207b8f1fd28adcdd5aafa7",
    "src/auto_research/reproductions/tapo_trajectory/objective.py": "49ab02ac758251acde90091a82b22576449d294a1f25cc3144ff1addd1132ae2",
    "src/auto_research/reproductions/tapo_trajectory/paper.yaml": "493e59ec67f96ee37ffecf6694dd5c4d8555849b8f756ccb63644da90bbb48d6",
    "src/auto_research/reproductions/upturn/__init__.py": "1e9793e9ea85e94144f7eac8fa141567d7f964438fa628788b668358154e196e",
    "src/auto_research/reproductions/upturn/adapter.py": "72e5da3cb70856944d0c9a25376f73d4a1d9f16213efd43e19b3acf2a584afc2",
    "src/auto_research/reproductions/upturn/algorithm.py": "421abd06cc492c39cf72e9b15c7f85c01e85fae0a7c26fb3053fbd19bc4897a4",
    "src/auto_research/reproductions/upturn/checkpoint.py": "fbcb7c830cf53d3130cfc06621f3178516aa72c7a939f4f765c3c7864b383787",
    "src/auto_research/reproductions/upturn/paper.yaml": "a1e335526f895b54e325331dc8354330f4a2d44f5f8c9c2a37f76ef2f7d589bc"
  },
  "provenance": {
    "commit": "32f873e7a81fa90e6508d5eaad5bf9b830b04139",
    "artifact_path": "docs/gpu-validations/progressive-contraction-a100-20261014-seeds42-44.json",
    "metrics_artifact": "docs/reproductions/2610.12933-progressive-contraction/metrics/public-a100-complete-seeds42-44.json",
    "execution_config_sha256": "506dce6568921b9c8fa75d969267537f9bb5c5c5b093a77b283616a2246524c5",
    "checkpoint_committed": false,
    "raw_results_sha256": {
      "42": "179bf4d73896074908a445f9e20e87ceba1e1c34c24927349fd19da312706e2d",
      "43": "e5cf7cc3e8d82fb0bd86ddf9df748225b6395f884c8d752128a992bdf582e8f2",
      "44": "1f2f014fdb944b8e6b7464a369799b40a236af866a43521f276395d17412bd69"
    }
  },
  "all_sanitized_commands": [
    [
      "python",
      "scripts/run_oct13_post_training.py",
      "--method",
      "progressive-contraction",
      "--config",
      "data/oct13_execution.json",
      "--seed",
      "42",
      "--output",
      "runs/complete-progressive-contraction-s42"
    ],
    [
      "python",
      "scripts/run_oct13_post_training.py",
      "--method",
      "progressive-contraction",
      "--config",
      "data/oct13_execution.json",
      "--seed",
      "43",
      "--output",
      "runs/complete-progressive-contraction-s43"
    ],
    [
      "python",
      "scripts/run_oct13_post_training.py",
      "--method",
      "progressive-contraction",
      "--config",
      "data/oct13_execution.json",
      "--seed",
      "44",
      "--output",
      "runs/complete-progressive-contraction-s44"
    ]
  ]
}