Multi-Hop Search-Augmented QA on 2Wiki (out-of-domain)
47.6AccuracyGEPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GEPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 47.6 | |
| GiGPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 43.6 | |
| Search-R1Type=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 40.1 | |
| GiGPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 37 | |
| GEPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 36.6 | |
| StepSearchType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 36.6 | |
| ZeroSearchType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 35.2 | |
| StepSearchType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 32 | |
| Search-R1Type=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 31.9 | |
| ZeroSearchType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 30 | |
| R1-InstructType=RL Training, Model=Qwen2.5-3B-Instruct2025.10 | 29.2 | |
| R1-InstructType=RL Training, Model=Qwen2.5-7B-Instruct2025.10 | 27.5 |