Calibration

MMVU → MathVision

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

MathVision is estimated from MMVU with a Michaelis–Menten curve fitted on 5 models measured on both: y = 2.0000·x / (0.97662 + x), R² = 0.73, cross-validated error 3.2 pp. It is used for 1 estimate.

Estimated model	MMVU	MathVision	Source
Kimi K2.5	80.4%	90.3%	estimated ± 3.2 pp, medium confidence