Multi-Hop Question Answering on HotpotQA (token-level F1)
62.6Token-level F1Oracle
Evaluation Results
| Method | Links | |
|---|---|---|
| OracleMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 62.6 | |
| COSEARCHMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 59.5 | |
| OracleMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 58.9 | |
| Fixed RankerMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 57.8 | |
| Search-R1Main agent backbone=Qwen2.5-7B-Instruct2026.04 | 57.6 | |
| ZeroSearchMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 57.5 | |
| Retrieval OnlyMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 57.2 | |
| COSEARCHMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 47.8 | |
| Fixed RankerMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 47.4 | |
| ZeroSearchMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 45.5 | |
| Retrieval OnlyMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 45.4 | |
| Search-R1Main agent backbone=Qwen2.5-3B-Instruct2026.04 | 45 | |
| RAGMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 37.1 | |
| RAGMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 37.1 | |
| Search-o1Main agent backbone=Qwen2.5-7B-Instruct2026.04 | 31.6 | |
| Direct InferenceMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 29.2 | |
| CoTMain agent backbone=Qwen2.5-7B-Instruct2026.04 | 24.4 | |
| Search-o1Main agent backbone=Qwen2.5-3B-Instruct2026.04 | 22.5 | |
| Direct InferenceMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 20.2 | |
| CoTMain agent backbone=Qwen2.5-3B-Instruct2026.04 | 19.7 |