{
  "schema_version": 2,
  "method": "crop",
  "manifest_ref": "agent-research:crop",
  "paper": {
    "key": "crop",
    "title": "CROP: Token-Efficient Reasoning in Large Language Models via Regularized Prompt Optimization"
  },
  "dataset": {
    "id": "synthetic:oct30-crop-v1",
    "revision": "sha256:6c73df58387b52fa4d73b68306b6f2991bc56096a35de30a90ae728fc50bbec9",
    "recipe": "fixed seeded formula/state/statistical contracts; no public capability dataset",
    "capability_benchmark": false
  },
  "seed": [
    42,
    43,
    44
  ],
  "seeds": [
    42,
    43,
    44
  ],
  "per_seed": [
    {
      "seed": 42,
      "diagnostic_only": true,
      "component": "crop_textual_prompt_control",
      "metrics": {
        "evaluated_prompt_count": 2,
        "brevity_callback_count": 2,
        "reference_normalized_score": 0.92,
        "reference_normalized_length": 0.4
      },
      "checks": {
        "strict_length_gate": true,
        "literal_gradient_concatenation": true,
        "validated_prompt_selected": true,
        "unvalidated_final_not_selected": true,
        "score_matches_equation4": true,
        "optimizer_reference_field_absent": true,
        "callback_doubles_not_capability": true
      },
      "execution_device": "cpu",
      "execution_accelerator": "CPU"
    },
    {
      "seed": 43,
      "diagnostic_only": true,
      "component": "crop_textual_prompt_control",
      "metrics": {
        "evaluated_prompt_count": 2,
        "brevity_callback_count": 2,
        "reference_normalized_score": 0.92,
        "reference_normalized_length": 0.4
      },
      "checks": {
        "strict_length_gate": true,
        "literal_gradient_concatenation": true,
        "validated_prompt_selected": true,
        "unvalidated_final_not_selected": true,
        "score_matches_equation4": true,
        "optimizer_reference_field_absent": true,
        "callback_doubles_not_capability": true
      },
      "execution_device": "cpu",
      "execution_accelerator": "CPU"
    },
    {
      "seed": 44,
      "diagnostic_only": true,
      "component": "crop_textual_prompt_control",
      "metrics": {
        "evaluated_prompt_count": 2,
        "brevity_callback_count": 2,
        "reference_normalized_score": 0.92,
        "reference_normalized_length": 0.4
      },
      "checks": {
        "strict_length_gate": true,
        "literal_gradient_concatenation": true,
        "validated_prompt_selected": true,
        "unvalidated_final_not_selected": true,
        "score_matches_equation4": true,
        "optimizer_reference_field_absent": true,
        "callback_doubles_not_capability": true
      },
      "execution_device": "cpu",
      "execution_accelerator": "CPU"
    }
  ],
  "metrics": {
    "brevity_callback_count": {
      "mean": 2,
      "std": 0.0,
      "n": 3
    },
    "evaluated_prompt_count": {
      "mean": 2,
      "std": 0.0,
      "n": 3
    },
    "reference_normalized_length": {
      "mean": 0.4,
      "std": 0.0,
      "n": 3
    },
    "reference_normalized_score": {
      "mean": 0.92,
      "std": 0.0,
      "n": 3
    }
  },
  "evaluation_tier": "l1_mechanism",
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42,
      43,
      44
    ],
    "formal_comparison": false,
    "claim_policy": "bounded component contracts only, not a capability or complete paper reproduction"
  },
  "diagnostic_only": true,
  "fixture_demo": true,
  "formal_comparison": false,
  "promotion_eligible": false,
  "evolve_integrated": false,
  "automatic_retry": false,
  "automatic_implementation_queue": false,
  "full_experiment_status": "not_reproduced_high_cost",
  "full_experiment_review_done": true,
  "provenance": {
    "commit": "af9b9245aa85b4cbeb3da27ca701cc01cb94da86",
    "artifact_path": "docs/agent-research/2604.14214-crop/metrics/component-cpu-seeds42-44.json",
    "accelerator": "CPU",
    "executed_source_sha256": {
      "scripts/run_oct30_components.py": "77a4a18b17587a04f06fda5b4d2d07998b50b8a7c670db61d594a12a94613379",
      "docs/research-audits/oct30-overlap-components/execution-spec.json": "0afad26f7fc9ee66a0febd18ea5cf7b9cb6ded10d4c09c6efc30be05594d54dc",
      "docs/research-audits/oct30-overlap-components/preflight.json": "193b954ae5600436cd55090b0a5e4e373f557963704e33ed70ecfd58cb99f4d5",
      "scripts/oct18_frozen_sources.py": "da08355049c1f6b5184ee659643c9be9decedf2041ef6a87a3dce2e4582e385a",
      "scripts/run_google_emnlp25_components.py": "bb336482f0927c7d1164d93534309a78d8258acaaab9f90549ef397fdf12a0d9",
      "src/auto_research/agent_research/crop.py": "6c73df58387b52fa4d73b68306b6f2991bc56096a35de30a90ae728fc50bbec9",
      "src/auto_research/agent_research/emma.py": "4c237168aff9f5de2821e800dbaff348d2b02233417683eb6c2acc3431187f07",
      "src/auto_research/agent_research/evalalloc.py": "e4debc6d0f1e4b2ee55703cec42278014a4fc10904a0ac8cd804983c612665d5",
      "src/auto_research/foundation_models/fold_bias.py": "c9d316b7e6310adb647460f1c7ef6c59b26bfcdd8543566fbd35947356b2f24c",
      "src/auto_research/foundation_models/merlin.py": "60267e6a6ff17dc96bb748737a6dbfae7fa234d2f17cdfe11f6dcc819b1a3836",
      "src/auto_research/foundation_models/swim.py": "4bedb2886bd22b76bc3225614d5028a77ce3e5eaf324735097ea4125cbcb1452",
      "src/auto_research/post_training/concur.py": "ea7de177dcc957e7c57b89b705b089ab8ce3e0be129b3be08ea8ab4c493f80b9",
      "src/auto_research/post_training/mpopd.py": "dcc3940889a0ea3c7522f9558493788126845aefe1dc74d26d321078d7edc229",
      "src/auto_research/reproductions/pinger_components.py": "0bc3fc9f8741a310fdfd1d41f6dc29b722144051c1ed4d5926e26c9402ef89db",
      "src/auto_research/reproductions/pinterest_rlbp.py": "3551a1ac348e0d70beff1fbd70950d9583f4913b9623bc455c1350e14a766d3b"
    },
    "raw_runner_artifact_sha256": "5680575049fa09e5ca738684184adc09618f04b6df3c4deb4c0c9730f6f009a7",
    "dataset_fingerprint": "sha256:0923fac79aeab78fb96c41e95824028ff7cd23b3fc2c839212c059650e66b5a4"
  }
}