Calibration

SimpleVQA → VideoMMMU

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

VideoMMMU is estimated from SimpleVQA with a Michaelis–Menten curve fitted on 5 models measured on both: y = 0.9479·x / (0.07077 + x), R² = 0.45, cross-validated error 2.2 pp. It is used for 1 estimate.

Estimated model	SimpleVQA	VideoMMMU	Source
Step 3.7 Flash	79.2%	87.0%	estimated ± 2.2 pp, low confidence