Calibration

ARC-AGI-1 → CritPt

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

CritPt is estimated from ARC-AGI-1 with a offset logistic curve fitted on 34 models measured on both: y = 0.0793 + (0.3500 − 0.0793) / (1 + exp(−26.41·(x − 0.9302))), R² = 0.69, cross-validated error 5.8 pp. It is used for 0 estimates.