Calibration

We-Math → MLVU (M-Avg)

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

MLVU (M-Avg) is estimated from We-Math with a offset logistic curve fitted on 6 models measured on both: y = 0.6288 + (0.8508 − 0.6288) / (1 + exp(−157.91·(x − 0.6891))), R² = 0.95, cross-validated error 3.4 pp. It is used for 0 estimates.

Fitted on

Model	We-Math	MLVU (M-Avg)
Claude Opus 4.5	70.0%	81.7%
Gemini 3 Pro	86.9%	83.0%
GPT-5.2	79.0%	85.6%
Kimi K2.5	84.7%	85.0%
LFM3-VL-2B	68.6%	71.3%
Qwen3.5 397B	87.9%	86.7%