Calibration

Video-MME (with subtitle) → CharXiv

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

CharXiv is estimated from Video-MME (with subtitle) with a linear curve fitted on 5 models measured on both: y = 4.3259·x + -2.9766, R² = 0.88, cross-validated error 2.7 pp. It is used for 1 estimate.

Estimated model	Video-MME (with subtitle)	CharXiv	Source
MiniMax M3	85.4%	71.8%	estimated ± 2.7 pp, low confidence