ResearchBenchmarksMulti-task Language Understanding on MMLU (dev)Follow86.4AccuracyGPT-465.225670.722876.2281.7172Jan 24, 2025Evaluation ResultsMethodMethodLinksAccuracyGPT-42025.0186.4GPT-3.52025.0170TAMA2025.0166.99base2025.0166.04