Multi-Hop Question Answering on MuSiQue (Exact Match %)
22.6Exact Match (%)StepSearch
Evaluation Results
| Method | Links | |
|---|---|---|
| StepSearchBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 22.6 | |
| IGRPOBackbone=Qwen2.5-7B-Instruct, Type=Tree-based2026.07 | 20.2 | |
| GiGPOBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 18.9 | |
| ZeroSearchBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 18.4 | |
| Tree-GRPOBackbone=Qwen2.5-7B-Instruct, Type=Tree-based2026.07 | 17.5 | |
| StepSearchBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 17.4 | |
| IGPOBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 16.8 | |
| Search-R1Backbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 14.6 | |
| IGRPOBackbone=Qwen2.5-3B-Instruct, Type=Tree-based2026.07 | 14.1 | |
| GiGPOBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 12.6 | |
| IGPOBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 12.3 | |
| Tree-GRPOBackbone=Qwen2.5-3B-Instruct, Type=Tree-based2026.07 | 11.9 | |
| Search-R1Backbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 10.3 | |
| ZeroSearchBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 9.8 | |
| R1-InstructBackbone=Qwen2.5-3B-Instruct, Type=RL Training2026.07 | 7.2 | |
| R1-InstructBackbone=Qwen2.5-7B-Instruct, Type=RL Training2026.07 | 6 |