Multi-hop Question Answering on HotpotQA (pass@1, pass@2)
65.9pass@1 Success RateEXG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EXGModel=Qwen-Max2026.05 | 65.9 | 68.4 | |
| EXG-SEModel=Qwen-Max2026.05 | 65.7 | 70.7 | |
| EXG-SE-RevModel=Qwen-Max2026.05 | 65.5 | 70.3 | |
| EXG-ReflexionModel=Qwen-Max2026.05 | 65.1 | 68.6 | |
| SE-Agent-RevModel=Qwen-Max2026.05 | 62.1 | 68 | |
| EXGModel=Qwen-Plus2026.05 | 61.8 | 64.2 | |
| ReflexionModel=Qwen-Max2026.05 | 61.6 | 68.6 | |
| EXG-SEModel=Qwen-Plus2026.05 | 61.3 | 68.1 | |
| EXG-SE-RevModel=Qwen-Plus2026.05 | 60.8 | 68 | |
| SE-AgentModel=Qwen-Max2026.05 | 60.7 | 66.9 | |
| SE-AgentModel=Qwen-Plus2026.05 | 60.3 | 67 | |
| EXG-ReflexionModel=Qwen-Plus2026.05 | 60.1 | 67.4 | |
| ReflexionModel=Qwen-Plus2026.05 | 59.7 | 66.2 | |
| SE-Agent-RevModel=Qwen-Plus2026.05 | 58.7 | 66.9 | |
| EXG-ReflexionModel=Qwen3-8B2026.05 | 46.2 | 55.7 | |
| EXGModel=Qwen3-8B2026.05 | 46 | 50 | |
| EXG-SE-RevModel=Qwen3-8B2026.05 | 45.8 | 50.9 | |
| EXG-SEModel=Qwen3-8B2026.05 | 44.5 | 51.6 | |
| ReflexionModel=Qwen3-8B2026.05 | 28.3 | 49.4 | |
| SE-AgentModel=Qwen3-8B2026.05 | 28.3 | 41 | |
| SE-Agent-RevModel=Qwen3-8B2026.05 | 27.5 | 34.4 |