Answer Harm on HotpotQA 104
1.9Harm Rate (HotpotQA 104)no-source variants
Evaluation results
| Method | Links | ||
|---|---|---|---|
| no-source variantsCampaign=Adaptive attacker, System / index=PPR / HotpotQA, N (cc)=104, judge=dual, 2s2026.06 | 1.9 | — | |
| AforgeCampaign=Adaptive attacker, System / index=PPR / HotpotQA, N (cc)=104, judge=dual, 2s2026.06 | 20 | 0 |