Single-Hop Question Answering on PopQA 2018 Wikipedia dump (dev)
47.2AccuracyMR-Search
Evaluation Results
| Method | Links | |
|---|---|---|
| MR-SearchBackbone=Qwen2.5-7b2026.03 | 47.2 | |
| MR-SearchBackbone=Qwen2.5-3b2026.03 | 46 | |
| Search-R1Backbone=Qwen2.5-3b2026.03 | 45.6 | |
| StepResearchBackbone=Qwen2.5-3b2026.03 | 45.6 | |
| Search-R1Backbone=Qwen2.5-7b2026.03 | 44.9 | |
| PPRMBackbone=Qwen2.5-7b2026.03 | 43.7 | |
| StepResearchBackbone=Qwen2.5-7b2026.03 | 43.1 | |
| ReSearchBackbone=Qwen2.5-3b2026.03 | 43 | |
| PPRMBackbone=Qwen2.5-3b2026.03 | 41.1 | |
| ReSearchBackbone=Qwen2.5-7b2026.03 | 39.1 | |
| Search-o1Backbone=Qwen2.5-3b2026.03 | 26.2 | |
| Direct InferenceBackbone=Qwen2.5-7b2026.03 | 14 | |
| Search-o1Backbone=Qwen2.5-7b2026.03 | 13.1 | |
| Direct InferenceBackbone=Qwen2.5-3b2026.03 | 10.8 |