Calibration

MMMLU → GPQA Diamond

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

GPQA Diamond is estimated from MMMLU with a Hill curve fitted on 5 models measured on both: y = 0.0000 + (1.2000 − 0.0000)·x^6.00 / (0.74931^6.00 + x^6.00), R² = 0.89, cross-validated error 2.1 pp. It is used for 0 estimates.