Calibration

V* → MathVision

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

MathVision is estimated from V* with a Michaelis–Menten curve fitted on 9 models measured on both: y = 1.3236·x / (0.49278 + x), R² = 0.85, cross-validated error 2.7 pp. It is used for 1 estimate.

Estimated model	V*	MathVision	Source
Step 3.7 Flash	95.3%	87.2%	estimated ± 2.7 pp, high confidence