Calibration
HealthBench Hard → AA-HLE
AA-HLE is estimated from HealthBench Hard with a Michaelis–Menten curve fitted on 11 models measured on both: y = 0.7442·x / (0.21855 + x), R² = 0.42, cross-validated error 11.0 pp. It is used for 0 estimates.
AA-HLE is estimated from HealthBench Hard with a Michaelis–Menten curve fitted on 11 models measured on both: y = 0.7442·x / (0.21855 + x), R² = 0.42, cross-validated error 11.0 pp. It is used for 0 estimates.