benchgap
Calibration

MMMU-Pro → RealWorldQA

RealWorldQA is estimated from MMMU-Pro with a linear curve fitted on 5 models measured on both: y = 0.2763·x + 0.6453, R² = 0.98, cross-validated error 4.8 pp. It is used for 2 estimates.

Estimated modelMMMU-ProRealWorldQASource
Interfaze Beta71.1%84.2%estimated ± 4.8 pp, medium confidence
Pareto 26.978.0%86.1%estimated ± 4.8 pp, medium confidence