Calibration

MMLU-Pro (Arcee) → CritPt

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

CritPt is estimated from MMLU-Pro (Arcee) with a offset logistic curve fitted on 6 models measured on both: y = 0.0079 + (0.0299 − 0.0079) / (1 + exp(−200.00·(x − 0.8565))), R² = 0.97, cross-validated error 0.6 pp. It is used for 0 estimates.