Question Answering on HotpotQA (F1, EM)
60.62F1 ScorePRAISE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PRAISEApproach=Ours2026.04 | 60.62 | 47.17 | |
| R1-SearcherApproach=Agentic Search2026.04 | 55.55 | 43.31 | |
| Search-R1Approach=Agentic Search2026.04 | 51.24 | 39.77 | |
| StepSearchApproach=Process Reward2026.04 | 50.02 | 38.31 | |
| TIPSApproach=Process Reward2026.04 | 45.98 | 35.34 | |
| Search-o1Approach=Agentic Search2026.04 | 44.86 | 34.07 | |
| ReasonRAGApproach=Process Reward2026.04 | 41.53 | 31.24 | |
| Naive RAGApproach=Standard2026.04 | 36.69 | 27.45 | |
| Direct LLMApproach=Standard2026.04 | 25.25 | 17.6 |