benchgap
Calibration

BabyVision → MathVision

MathVision is estimated from BabyVision with a Hill curve fitted on 5 models measured on both: y = 0.8644 + (1.2000 − 0.8644)·x^4.41 / (1.03779^4.41 + x^4.41), R² = 0.99, cross-validated error 0.5 pp. It is used for 2 estimates.

Estimated modelBabyVisionMathVisionSource
GLM-5.3-Flash53.4%88.1%estimated ± 0.5 pp, medium confidence
Muse Spark 1.176.3%93.3%estimated ± 0.5 pp, medium confidence