| { | |
| "canonical_text_language": "en", | |
| "configs": { | |
| "evaluation_scenarios_960": { | |
| "rows": 960, | |
| "split": "test", | |
| "jsonl_path": "scenarios/evaluation_scenarios_960.jsonl", | |
| "parquet_path": "scenarios/evaluation_scenarios_960/test.parquet" | |
| }, | |
| "scenarios_2773": { | |
| "rows": 2773, | |
| "split": "corpus", | |
| "jsonl_path": "scenarios/scenarios_2773.jsonl", | |
| "parquet_path": "scenarios/scenarios_2773/corpus.parquet" | |
| } | |
| }, | |
| "construction_id_policy": "construction_scenario_id preserves the original generated scenario artifact ID; it is not the image join key.", | |
| "duplicate_text_policy": "Preserve every source-scenario ID; identical text may occur under distinct IDs.", | |
| "fields": { | |
| "concept": "string", | |
| "construction_scenario_id": "string", | |
| "is_evaluation": "bool", | |
| "scenario": "string", | |
| "scenario_id": "string", | |
| "value_dimension": "string", | |
| "value_endpoints": "list[struct{id: string, description: string}]" | |
| }, | |
| "join_key": "scenario_id joins the image dataset scenario_id exactly", | |
| "schema_version": "1.0", | |
| "description": "Canonical English scenarios after orientation screening; evaluation is a subset of the retained corpus." | |
| } | |