Question Answering on Humanity's Last Exam (HLE) curated 649-question subset (test)
54.3AccuracyFUSE
Evaluation Results
| Method | Links | |
|---|---|---|
| FUSEBest-of-k=502026.04 | 54.3 | |
| Oracle Best Verifier (GPT-5.2 High)Best-of-k=502026.04 | 53.5 | |
| Logistic RegressionBest-of-k=502026.04 | 53.4 | |
| Pass@1Best-of-k=502026.04 | 52.1 | |
| Naive BayesBest-of-k=502026.04 | 52 | |
| Naive EnsembleBest-of-k=502026.04 | 51.4 | |
| WEAVERBest-of-k=502026.04 | 51.2 |