General Question Answering on PopQA out-of-domain (val test)
50.1Exact Match (EM)Search-R2
Evaluation Results
| Method | Links | |
|---|---|---|
| Search-R2Backbone=Qwen2.5-32B2026.02 | 50.1 | |
| Search-R1Backbone=Qwen2.5-32B2026.02 | 47 | |
| Search-R2Backbone=Qwen3-8B2026.02 | 46.6 | |
| Search-R1Backbone=Qwen3-8B2026.02 | 41.8 | |
| Search-R2Backbone=Qwen2.5-7B2026.02 | 41 | |
| RAGBackbone=Qwen2.5-7B2026.02 | 39.2 | |
| Search-R1Backbone=Qwen2.5-7B2026.02 | 38.8 | |
| Rejection SamplingBackbone=Qwen2.5-7B2026.02 | 38 | |
| IRCoTBackbone=Qwen2.5-7B2026.02 | 30.1 | |
| R1-baseBackbone=Qwen2.5-7B2026.02 | 20.2 | |
| R1-instructBackbone=Qwen2.5-7B2026.02 | 19.9 | |
| Direct InferenceBackbone=Qwen2.5-7B2026.02 | 14 | |
| Search-o1Backbone=Qwen2.5-7B2026.02 | 13.1 | |
| SFTBackbone=Qwen2.5-7B2026.02 | 12.1 | |
| CoTBackbone=Qwen2.5-7B2026.02 | 5.4 |