Multi-hop Question Answering on HotpotQA (Evidence score)
33.2Evidence ScoreEVE-Agent
Evaluation Results
| Method | Links | |
|---|---|---|
| EVE-AgentSearch=true2026.05 | 33.2 | |
| Initial (no search)Search=false2026.05 | 24.7 | |
| Dr. ZeroSearch=true2026.05 | 20.9 | |
| Initial (search)Search=true2026.05 | 19.9 |