Multi-Hop Question Answering on Bamboogle (Exact Match (EM))
73.8Exact Match (EM)SkillRL
Evaluation Results
| Method | Links | |
|---|---|---|
| SkillRLEvaluation Protocol=Zero-shot2026.05 | 73.8 | |
| SKILLGRAPHEvaluation Protocol=Zero-shot2026.05 | 72.6 | |
| IGRPOBackbone=Qwen2.5-7B-Instruct, Type=Tree-based2026.07 | 71.8 | |
| GiGPOBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 68.9 | |
| IGRPOBackbone=Qwen2.5-3B-Instruct, Type=Tree-based2026.07 | 65.7 | |
| IGPOBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 64.5 | |
| GiGPOBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 64.1 | |
| EvolveREvaluation Protocol=Zero-shot2026.05 | 54.4 | |
| Tree-GRPOBackbone=Qwen2.5-7B-Instruct, Type=Tree-based2026.07 | 41.6 | |
| StepSearchBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 40 | |
| IGPOBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 39.9 | |
| Search-R1Evaluation Protocol=Zero-shot2026.05 | 36.8 | |
| Search-R1Backbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 36.8 | |
| StepSearchBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 34.4 | |
| Search-o1Evaluation Protocol=Zero-shot2026.05 | 30.4 | |
| R1-InstructBackbone=Qwen2.5-3B-Instruct, Type=RL Training2026.07 | 29.3 | |
| Tree-GRPOBackbone=Qwen2.5-3B-Instruct, Type=Tree-based2026.07 | 28 | |
| ZeroSearchEvaluation Protocol=Zero-shot2026.05 | 27.8 | |
| ZeroSearchBackbone=Qwen2.5-7B-Instruct, Type=Chain-based2026.07 | 27.8 | |
| Search-R1Backbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 26.4 | |
| CoTEvaluation Protocol=Zero-shot2026.05 | 24 | |
| R1-InstructBackbone=Qwen2.5-7B-Instruct, Type=RL Training2026.07 | 19.2 | |
| RAGEvaluation Protocol=Zero-shot2026.05 | 16.8 | |
| Qwen2.5-7B-InstructEvaluation Protocol=Zero-shot2026.05 | 14.4 | |
| ZeroSearchBackbone=Qwen2.5-3B-Instruct, Type=Chain-based2026.07 | 11.1 |