| { |
| "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" |
| } |
|
|
