Multi-Hop Question Answering on HotpotQA (OSR%)
33.9OSR ScoreStepSearch-base
Evaluation Results
| Method | Links | |
|---|---|---|
| StepSearch-baseBackbone=Qwen2.5-3B2026.04 | 33.9 | |
| Search-R1-baseBackbone=Qwen2.5-7B2026.04 | 32.4 | |
| StepSearch-instructBackbone=Qwen2.5-7B2026.04 | 27.3 | |
| StepSearch-baseBackbone=Qwen2.5-7B2026.04 | 23.8 | |
| Search-R1-instructBackbone=Qwen2.5-7B2026.04 | 23.4 | |
| StepSearch-instructBackbone=Qwen2.5-3B2026.04 | 18.1 | |
| Search-R1-instructBackbone=Qwen2.5-3B2026.04 | 13.9 | |
| HIPRAG-instructBackbone=Qwen2.5-7B2026.04 | 12.5 | |
| Search-R1-baseBackbone=Qwen2.5-3B2026.04 | 9.2 | |
| AutoSearchBackbone=Qwen2.5-7B2026.04 | 5.78 | |
| HIPRAG-baseBackbone=Qwen2.5-3B2026.04 | 2.6 | |
| HIPRAG-instructBackbone=Qwen2.5-3B2026.04 | 1.7 | |
| AutoSearchBackbone=Qwen2.5-3B2026.04 | 0.9 |