Datasets:
| license: cc-by-4.0 | |
| tags: | |
| - data-use | |
| - ner | |
| - validation | |
| configs: | |
| - config_name: default | |
| data_files: | |
| - split: train | |
| path: train.jsonl | |
| # fcv-validations | |
| Luna-judged (`gpt-5.6-luna`) verdicts for FCV data-mention extractions. | |
| One row per extracted mention. Fields: `doc_id` (= extraction `corpus_id`), | |
| `page`, `chunk`, `start`, `end`, `mention_name`, `extractor_specificity`, and the | |
| judge fields `valid_data_mention` (bool), `specificity`, `typology`, `usage`. | |
| Rows the judge could not parse carry `judge_error` instead of the judge fields. | |
| Configs present (by `doc_id` prefix): `fcv_pads_east_asia`, `jdc_operational`, | |
| `refugee_pads`, `reliefweb`. | |