·
13 commits
to main
since this release
Hard518 public release package.
Includes:
- data/hard_518_queries.csv (fused_query, blurred_query)
- data/gold_public_hard518.jsonl (gold nuggets)
- results/candidates/*.jsonl (baseline candidate answers)
- results/clarify_only/*.csv (questions/answers/rewrites)
- code/ evaluation scripts (eval_gold.py, etc.)
Quick start:
export QIANFAN_API_KEY="YOUR_KEY"
python code/eval_gold.py --gold_a data/gold_public_hard518.jsonl --gold_b results/candidates/ebk1__cand_hard518.jsonl --out_dir outputs/ebk1