{
  "checkpoint_report_binding": "passed",
  "checkpoint_sha256": "7f21e80e61d16a71735163ef555d3009afb0c98da74c48e29df08606973cc55e",
  "code_hash_end": "08cbbaac2aa8f60af5c43af7a83410e9690b44c5e4d162fbe1e5766885952fce",
  "code_hash_start": "08cbbaac2aa8f60af5c43af7a83410e9690b44c5e4d162fbe1e5766885952fce",
  "dataset_id": "flip2-rhomax-by-wild-type",
  "environment": {
    "fair-esm": "2.0.0",
    "numpy": "1.26.4",
    "scikit-learn": "1.9.1",
    "scipy": "1.17.1",
    "torch": "2.2.2"
  },
  "esm2_load_inference_fit_evaluate_seconds": 64.55955554200045,
  "integrity_reaudit": {
    "date": "2026-09-21",
    "method": "--verify-existing on original outputs and checkpoint; no model rerun",
    "script_sha256": "8fa5094356b3dcc59c048d0a2a28cf90e5138fae8d9712b384f5af608c08e4a5",
    "timing_policy": "Original execution timing and duplicate-evidence annotations preserved"
  },
  "limitations": [
    "No confidence intervals",
    "Single task and split",
    "Unreported pretraining overlap",
    "No container validation",
    "No upload/publication performed by this script"
  ],
  "prepared_sha256": "f417d43497e0c8c453bd7c66cfea05959b4d6d47462283096e42a2dae8b72419",
  "review_method": "automated independent metric recomputation from local predictions",
  "runs": [
    {
      "evidence_origin": "Rewire local model execution, not paper reproduction",
      "independent_metric_recomputation": "passed",
      "ndcg": 0.9072580204736261,
      "prediction_digest_verification": "passed",
      "predictions_sha256": "96e8a0c6663dd0a471b984fb86506e483d252a6036a9617cf5f4487664274abd",
      "prepared_and_code_binding": "passed",
      "run": "esm2",
      "scored": 184,
      "spearman": -0.2217595033467806,
      "tolerance": 1e-12
    },
    {
      "evidence_origin": "Rewire local model execution, not paper reproduction",
      "independent_metric_recomputation": "passed",
      "ndcg": 0.9206667522227658,
      "prediction_digest_verification": "passed",
      "predictions_sha256": "599dca4e6fa643a7cdfc96ac339033f61e5846ce225af17ac2ac3917ecc14fac",
      "prepared_and_code_binding": "passed",
      "relationship_to_existing_evidence": {
        "classification": "validation_only_not_additional_independent_evidence",
        "original_receipt": "../rhomax/verification.json",
        "predictions_match_exactly": true,
        "submit_again": false
      },
      "run": "training-mean-v1",
      "scored": 184,
      "spearman": null,
      "tolerance": 1e-12
    },
    {
      "evidence_origin": "Rewire local model execution, not paper reproduction",
      "independent_metric_recomputation": "passed",
      "ndcg": 0.9548154942827918,
      "prediction_digest_verification": "passed",
      "predictions_sha256": "0a214b7dbad99d3684353cf3d56c5c57bae1e2dfa67eb88f67e3f32c66313dd7",
      "prepared_and_code_binding": "passed",
      "relationship_to_existing_evidence": {
        "classification": "validation_only_not_additional_independent_evidence",
        "original_receipt": "../rhomax/verification.json",
        "predictions_match_exactly": true,
        "submit_again": false
      },
      "run": "protein-composition-probe-v1",
      "scored": 184,
      "spearman": 0.41798958279508075,
      "tolerance": 1e-12
    }
  ],
  "schema_version": "1.0",
  "scope": "one complete Rhomax split; no suite aggregate",
  "source_verification": "pinned_source_bytes",
  "split_counts": {
    "test": 184,
    "train": 584,
    "validation": 116
  },
  "timing_scope": "elapsed is null when auditing existing outputs; execution.inference_and_fit_seconds in the original report covers encoding plus fitting, excludes checkpoint load/acquisition"
}