Calibration

NOVA-63 → MMLU-ProX

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

MMLU-ProX is estimated from NOVA-63 with a Michaelis–Menten + offset curve fitted on 7 models measured on both: y = 0.0054 + 1.6453·x / (0.54905 + x), R² = 0.52, cross-validated error 1.4 pp. It is used for 0 estimates.