{
  "blinding": "second rater saw raw RL export and local AIGR reference only",
  "both_axes_exact_agreement": 0.55,
  "completeness": {
    "exact_agreement": 0.55,
    "first_distribution": {
      "1": 4,
      "2": 8,
      "3": 4,
      "4": 4
    },
    "first_mean": 2.4,
    "mean_absolute_error": 0.5,
    "pearson": 0.77555,
    "quadratic_weighted_kappa": 0.74359,
    "second_distribution": {
      "1": 3,
      "2": 6,
      "3": 6,
      "4": 4,
      "5": 1
    },
    "second_mean": 2.7,
    "spearman": 0.758789,
    "within_one_agreement": 0.95
  },
  "correctness": {
    "exact_agreement": 0.8,
    "first_distribution": {
      "1": 4,
      "2": 4,
      "3": 4,
      "4": 4,
      "5": 4
    },
    "first_mean": 3.0,
    "mean_absolute_error": 0.2,
    "pearson": 0.959645,
    "quadratic_weighted_kappa": 0.95,
    "second_distribution": {
      "1": 3,
      "2": 4,
      "3": 4,
      "4": 4,
      "5": 5
    },
    "second_mean": 3.2,
    "spearman": 0.960878,
    "within_one_agreement": 1.0
  },
  "n": 20,
  "sample_salt": "argo-audit-2026-07-11",
  "sampling": "four SHA-256-selected genes per first-rater correctness stratum"
}