{
  "authored_semantic_bases": 24,
  "authored_supported_bases": 10,
  "authored_supported_bases_correct": 8,
  "authored_variants_per_base": 5,
  "banks": {
    "final_challenge": {
      "bank_file_sha256": "edf44de829ea42599e70c0cd8a0a8f925c741d7106cf42539175f11c59e86883",
      "full_decisions_correct": 97,
      "negative": 70,
      "negative_false_accepts": 13,
      "rule_supported_exact": 41,
      "supported": 50,
      "supported_exact": 40,
      "supported_execution_correct": 40,
      "supported_rejected": 10,
      "total": 120
    },
    "fresh": {
      "bank_file_sha256": "3da8357ffc18e3dd836665758d747db61708314eccc63f735423bc5d7f6b8670",
      "full_decisions_correct": 1192,
      "negative": 575,
      "negative_false_accepts": 5,
      "rule_supported_exact": 555,
      "supported": 625,
      "supported_exact": 622,
      "supported_execution_correct": 622,
      "supported_rejected": 3,
      "total": 1200
    },
    "transfer": {
      "bank_file_sha256": "20b9c3c7e2f851f69de85f163e57ac2c6a564b3dfd2d8304ff3f68261cff80aa",
      "full_decisions_correct": 1077,
      "negative": 588,
      "negative_false_accepts": 8,
      "rule_supported_exact": 503,
      "supported": 612,
      "supported_exact": 497,
      "supported_execution_correct": 497,
      "supported_rejected": 115,
      "total": 1200
    }
  },
  "checkpoint_sha256": "a6b944d56580b64dfb526da5b3888a137b6783ea29d8a719a0fea3f4e3f7ee76",
  "historical_evaluation_record_sha256": "c6644ab82f5d7411b92b0bcb5c387412357a0caf829c76b15fb26b90bfd57c38",
  "historical_training_record_sha256": "871670888d326374c211c511cfb6816779ae899774f729cf6bc60935a6516a0a",
  "limitations": [
    "Synthetic/AI-authored data, not real-user or population evidence.",
    "Authored variants are correlated; 120 requests are not 120 independent users.",
    "All banks are exposed now; not future held-out evidence.",
    "Dataset hashes identify source bytes, not proof of label correctness.",
    "Compiler execution used fake tools, not a security boundary.",
    "No attention-candidate results are included."
  ],
  "parameters": 67485,
  "status": "historical_exposed_synthetic_evaluation_not_new_measurement",
  "training": {
    "epochs_completed": 55,
    "initialization": "random",
    "seed": 715,
    "selected_epoch": 12,
    "selected_validation_exact": 0.990625,
    "training_sha256": "ed2ff9f0275c86958c475a89ec8e90664b954e0408eaefc356aa5ffb81c3b33a"
  }
}