Blame for DATASET_STATISTICS.json

{
  "format": "parquet + lossless JSONL mirrors",
  "record_counts": {
    "core_evaluation": 324,
    "probe_evaluation": 120,
    "strong_evaluation": 192,
    "unified_evaluation": 32,
    "warmup_usages": 41,
    "contexts": 74
  },
  "evaluation_records": 668,
  "distinct_future_teacher_cases": 148,
  "warmup_usage_records": 41,
  "distinct_warmup_teacher_cases": 21,
  "audited_risk_records": 709,
  "evaluation_certified": 661,
  "evaluation_refused": 7,
  "observed_false_certificates": 0,
  "repeated_cases_are_not_independent": true,
  "contexts_are_accounting_tables_not_extra_teacher_examples": true
}