{
  "model": "E-V3.3-frozen-mechanism-order",
  "benchmark_type": "retrospective_frozen_constraints_not_independent_prospective",
  "benchmark_items": 26,
  "true_score": 26,
  "failed_true_items": [],
  "randomized_mappings": 40320,
  "null_mean": 7.071428571428571,
  "null_sd": 3.156006043983418,
  "max_null_score": 26,
  "n_at_or_above_true": 1,
  "exact_tail_fraction": 2.48015873015873e-05,
  "distribution": {
    "1": 326,
    "2": 1631,
    "3": 2992,
    "4": 4100,
    "5": 4710,
    "6": 5085,
    "7": 4910,
    "8": 4432,
    "9": 3654,
    "10": 2865,
    "11": 1997,
    "12": 1380,
    "13": 914,
    "14": 567,
    "15": 323,
    "16": 175,
    "17": 109,
    "18": 60,
    "19": 40,
    "20": 25,
    "21": 13,
    "22": 5,
    "23": 4,
    "24": 2,
    "26": 1
  },
  "interpretation": "The frozen author-to-dynamic-state mapping is uniquely best under this retrospective constraint set. Because several constraints come from earlier model-development/reverse-test records, this is a stress test against label permutation, not an independent confirmatory p-value."
}