Calibration

MMLU-Pro → GPQA Diamond

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

GPQA Diamond is estimated from MMLU-Pro with a Hill curve fitted on 22 models measured on both: y = 0.2417 + (1.2000 − 0.2417)·x^4.92 / (0.75565^4.92 + x^4.92), R² = 0.96, cross-validated error 5.1 pp. It is used for 0 estimates.