Calibration
AA-GPQA Diamond → AA-Omniscience Accuracy
AA-Omniscience Accuracy is estimated from AA-GPQA Diamond with a offset logistic curve fitted on 169 models measured on both: y = 0.1458 + (1.2000 − 0.1458) / (1 + exp(−12.47·(x − 0.9961))), R² = 0.64, cross-validated error 7.6 pp. It is used for 0 estimates.