benchgap
Calibration

AA-Omniscience Accuracy → AA-GPQA Diamond

AA-GPQA Diamond is estimated from AA-Omniscience Accuracy with a Michaelis–Menten curve fitted on 169 models measured on both: y = 1.0574·x / (0.08190 + x), R² = 0.63, cross-validated error 10.8 pp. It is used for 0 estimates.