Calibration
BioBasicsBench → HealthBench Professional
HealthBench Professional is estimated from BioBasicsBench with a Michaelis–Menten curve fitted on 5 models measured on both: y = 0.8554·x / (0.20223 + x), R² = 0.70, cross-validated error 6.1 pp. It is used for 0 estimates.
Fitted on
| Model | BioBasicsBench | HealthBench Professional |
|---|---|---|
| GPT-5.4 | 33.2% | 48.1% |
| GPT-5.6 Sol | 45.6% | 60.5% |
| GPT-6.1 Sol | 62.6% | 64.2% |
| GPT-6 Astra | 67.0% | 64.7% |
| GPT-6 Luna | 37.6% | 60.8% |