{
  "version": "textinsightbench",
  "evaluation_unit": "evidence-backed downstream finding",
  "reference_availability": "public; disclose reference access during development and evaluation",
  "difficulty_profiles": {
    "standard": "Original discovery and evidence requirements; finding-quality scoring.",
    "hard": "Stratification, dominant-stratum sensitivity, support concentration and stronger evidence; finding-quality-robustness scoring. Same corpus and base discovery references, not independent confirmation."
  },
  "task_families": {
    "group_difference": 20,
    "temporal_change": 15,
    "compound_association": 15
  },
  "tracks": {
    "task_only": "Use each supplied task corpus.",
    "unlabeled_pool": "Learn from the released pool before evaluation; corpus-local analysis is allowed."
  },
  "reference_scope": "Non-exhaustive organizer reference set. Supported novel findings can earn full quality credit.",
  "cross_task_adaptation": "Freeze global prompts, learned parameters and thresholds before evaluation. Do not transfer evaluation feedback or task-fitted state to subsequent tasks.",
  "sampling": "Finite metadata-selected corpora; estimates describe supplied reports, not population incidence.",
  "independence": "Tasks share sources and entities. Document separation does not establish statistical independence.",
  "security": "Corpus content is untrusted input. Use separate execution environments for participants and organizers.",
  "scoring_version": "finding-quality"
}