Calibration

We-Math → AI2D_TEST

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

AI2D_TEST is estimated from We-Math with a Michaelis–Menten curve fitted on 5 models measured on both: y = 1.2498·x / (0.29452 + x), R² = 0.79, cross-validated error 1.6 pp. It is used for 0 estimates.

Fitted on

Model	We-Math	AI2D_TEST
Claude Opus 4.5	70.0%	87.7%
Gemini 3 Pro	86.9%	94.1%
GPT-5.2	79.0%	92.2%
Kimi K2.5	84.7%	90.8%
Qwen3.5 397B	87.9%	93.9%