Calibration
HealthBench Professional → AA-GPQA Diamond
AA-GPQA Diamond is estimated from HealthBench Professional with a offset logistic curve fitted on 7 models measured on both: y = 0.9145 + (0.9617 − 0.9145) / (1 + exp(−200.00·(x − 0.6045))), R² = 0.75, cross-validated error 2.1 pp. It is used for 4 estimates.
| Estimated model | HealthBench Professional | AA-GPQA Diamond | Source |
|---|---|---|---|
| Claude Opus 5.5 | 65.6% | 96.2% | estimated ± 2.1 pp, low confidence |
| Claude Sonnet 5.5 | 69.2% | 96.2% | estimated ± 2.1 pp, low confidence |
| Grok 4.7 | 56.7% | 91.5% | estimated ± 2.1 pp, medium confidence |
| Ling 3.1 Flash | 65.4% | 96.2% | estimated ± 2.1 pp, low confidence |