Multi-hop QA on 2WikiMHQA Wiki18 index (test)
55.1Pass@4Qwen3-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-8BSampling Method=DivInit (DI)2026.06 | 55.1 | |
| Gemma3-12BSampling Method=DivInit (DI)2026.06 | 53.9 | |
| Gemma3-4BSampling Method=DivInit (DI)2026.06 | 52.2 | |
| Gemma3-12BSampling Method=Standard parallel sampling (S)2026.06 | 52 | |
| Gemma3-12BSampling Method=DivInit (DI)2026.06 | 50.2 | |
| Qwen3-4BSampling Method=DivInit (DI)2026.06 | 49 | |
| Qwen3-8BSampling Method=Standard parallel sampling (S)2026.06 | 46.3 | |
| Qwen3-8BSampling Method=DivInit (DI)2026.06 | 46 | |
| Gemma3-12BSampling Method=Standard parallel sampling (S)2026.06 | 45 | |
| Gemma3-4BSampling Method=Standard parallel sampling (S)2026.06 | 42.8 | |
| Qwen3-4BSampling Method=Standard parallel sampling (S)2026.06 | 41.9 | |
| Qwen3-1.7BSampling Method=DivInit (DI)2026.06 | 41.5 | |
| Qwen3-8BSampling Method=Standard parallel sampling (S)2026.06 | 38.6 | |
| Qwen3-1.7BSampling Method=Standard parallel sampling (S)2026.06 | 37.6 | |
| Qwen3-4BSampling Method=DivInit (DI)2026.06 | 36.6 | |
| Gemma3-4BSampling Method=DivInit (DI)2026.06 | 34 | |
| Qwen3-4BSampling Method=Standard parallel sampling (S)2026.06 | 29.5 | |
| Gemma3-4BSampling Method=Standard parallel sampling (S)2026.06 | 28 | |
| Qwen3-1.7BSampling Method=DivInit (DI)2026.06 | 27.8 | |
| Qwen3-1.7BSampling Method=Standard parallel sampling (S)2026.06 | 25 |