benchmark_category,benchmarks,n,p
Math,"AIME 2024, AIME 2025, GPQA Diamond, HMMT 2025, Humanity's Last Exam, MATH, MMLU, MMLU-Pro, MMMLU, MMMU",10,0.4166666666666667
Reasoning and knowledge,"GPQA Diamond, Humanity's Last Exam, MMLU, MMLU-Pro, MMMLU, MMMU",6,0.25
Coding,"Aider Polyglot, LiveCodeBench, SWE-bench Verified",3,0.125
Audio-visual understanding,"Humanity's Last Exam, MMMU",2,0.08333333333333333
Factuality,SimpleQA,1,0.041666666666666664
Instruction following,IF Eval,1,0.041666666666666664
Multilingual performance,MMMLU,1,0.041666666666666664