benchgap
Calibration

Video-MME (with subtitle) → RealWorldQA

RealWorldQA is estimated from Video-MME (with subtitle) with a Michaelis–Menten + offset curve fitted on 9 models measured on both: y = 0.0000 + 2.0000·x / (1.20407 + x), R² = 0.68, cross-validated error 2.0 pp. It is used for 1 estimate.

Estimated modelVideo-MME (with subtitle)RealWorldQASource
MiniMax M385.4%83.0%estimated ± 2.0 pp, high confidence