Question Answering on PopQA (EM)
39.7EMSearch-R1
Evaluation results
| Method | Links | |
|---|---|---|
| Search-R1Type=RL Training2026.01 | 39.7 | |
| ACRType=SFT, Reasoner=Qwen-2.5-7B-Instruct2026.01 | 36.04 | |
| IRCoTType=Prompt2026.01 | 30.1 | |
| R1-InstructionType=RL Training2026.01 | 17.1 | |
| Direct inferenceType=Prompt2026.01 | 14 | |
| SFTType=SFT2026.01 | 12.1 | |
| CoTType=Prompt2026.01 | 5.4 |