Open-Web Reasoning on Humanity’s Last Exam (HLE) Open-web via SERPER (test)
14.8Pass@4Gemma3-12B
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemma3-12BSampling Method=DivInit (DI)2026.06 | 14.8 | |
| Qwen3-4BSampling Method=DivInit (DI)2026.06 | 14.3 | |
| Qwen3-8BSampling Method=DivInit (DI)2026.06 | 14.1 | |
| Gemma3-12BSampling Method=Standard parallel sampling (S)2026.06 | 12.7 | |
| Qwen3-8BSampling Method=Standard parallel sampling (S)2026.06 | 10 | |
| Qwen3-4BSampling Method=Standard parallel sampling (S)2026.06 | 9.7 |