Calibration
Humanity's Last Exam → CritPt
CritPt is estimated from Humanity's Last Exam with a Hill curve fitted on 27 models measured on both: y = 0.0041 + (0.3628 − 0.0041)·x^6.00 / (0.42882^6.00 + x^6.00), R² = 0.90, cross-validated error 3.5 pp. It is used for 3 estimates.
| Estimated model | Humanity's Last Exam | CritPt | Source |
|---|---|---|---|
| Claude Fable 5.1 (high with fallback) | 55.9% | 30.2% | estimated ± 3.5 pp, high confidence |
| Claude Opus 5.5 (high with fallback) | 55.6% | 30.0% | estimated ± 3.5 pp, high confidence |
| Claude Fable 5 (with fallback) | 55.5% | 30.0% | estimated ± 3.5 pp, high confidence |