Updates validation rows only. Train and holdout files unchanged. Validation labels remain agent-authored (owner_reviewed=false); scope and candidate-selection caveat recorded in README and split_stats.json.
28 days ago
Describe local LLM human-in-loop annotation workflow
Holdout-only post-hoc passage audit of gliner_agents_v3 disagreement candidates. Train and validation JSONL are intentionally unchanged. Labels remain agent-adjudicated (owner_reviewed=false); this is not an independent owner-gold benchmark.
29 days ago
raw-extraction build (model predictions, no judge)
