Multi-Hop Question Answering on Musique 2018 Wikipedia dump (dev)
22.1AccuracyMR-Search
Evaluation Results
| Method | Links | |
|---|---|---|
| MR-SearchBackbone=Qwen2.5-7b2026.03 | 22.1 | |
| StepResearchBackbone=Qwen2.5-7b2026.03 | 20.5 | |
| Search-R1Backbone=Qwen2.5-7b2026.03 | 18.1 | |
| ReSearchBackbone=Qwen2.5-7b2026.03 | 16.6 | |
| MR-SearchBackbone=Qwen2.5-3b2026.03 | 16.5 | |
| PPRMBackbone=Qwen2.5-7b2026.03 | 14.7 | |
| PPRMBackbone=Qwen2.5-3b2026.03 | 12.7 | |
| StepResearchBackbone=Qwen2.5-3b2026.03 | 10.5 | |
| Search-R1Backbone=Qwen2.5-3b2026.03 | 7.7 | |
| ReSearchBackbone=Qwen2.5-3b2026.03 | 7.4 | |
| Search-o1Backbone=Qwen2.5-7b2026.03 | 5.8 | |
| Search-o1Backbone=Qwen2.5-3b2026.03 | 5.4 | |
| Direct InferenceBackbone=Qwen2.5-7b2026.03 | 3.1 | |
| Direct InferenceBackbone=Qwen2.5-3b2026.03 | 2 |