Answer Harm on HotpotQA 37k-node
0Blind Harm Countpre-selection re-derivation
Evaluation results
| Method | Links | ||
|---|---|---|---|
| pre-selection re-derivationCampaign=Necessity panel, System / index=PPR / HotpotQA 37k-node2026.06 | 0 | — | |
| AUTHSELECTCampaign=Live external action, System / index=PPR / HotpotQA 37k-node2026.06 | 0 | 0 | |
| PPRCampaign=Necessity panel, System / index=PPR / HotpotQA 37k-node, N (cc)=615 (900), judge=dual, 3s2026.06 | 18 | — | |
| Published post-selection defensesCampaign=Necessity panel, System / index=PPR / HotpotQA 37k-node2026.06 | 18 | — | |
| PPRCampaign=Live external action, System / index=PPR / HotpotQA 37k-node, N (cc)=499 (900), judge=dual, 3s2026.06 | 18 | 5.61 |