Multi-Hop Question Answering on 2WikiMultiHopQA out-of-domain (val test)
51.7Exact Match (EM)Search-R2
Evaluation Results
| Method | Links | |
|---|---|---|
| Search-R2Backbone=Qwen2.5-32B2026.02 | 51.7 | |
| Search-R1Backbone=Qwen2.5-32B2026.02 | 46.2 | |
| Search-R2Backbone=Qwen3-8B2026.02 | 40.5 | |
| Search-R2Backbone=Qwen2.5-7B2026.02 | 35.8 | |
| Search-R1Backbone=Qwen3-8B2026.02 | 35.5 | |
| Search-R1Backbone=Qwen2.5-7B2026.02 | 29.7 | |
| Rejection SamplingBackbone=Qwen2.5-7B2026.02 | 29.6 | |
| R1-instructBackbone=Qwen2.5-7B2026.02 | 29.2 | |
| R1-baseBackbone=Qwen2.5-7B2026.02 | 27.3 | |
| SFTBackbone=Qwen2.5-7B2026.02 | 25.9 | |
| Direct InferenceBackbone=Qwen2.5-7B2026.02 | 25 | |
| RAGBackbone=Qwen2.5-7B2026.02 | 23.5 | |
| Search-o1Backbone=Qwen2.5-7B2026.02 | 17.6 | |
| IRCoTBackbone=Qwen2.5-7B2026.02 | 14.9 | |
| CoTBackbone=Qwen2.5-7B2026.02 | 11.1 |