ResearchBenchmarksMassive Multitask Language Understanding on MMLU hiFollow24.8MMLU hi AccuracyFXT λ324.2824.41524.5524.685Jul 17, 2025Evaluation ResultsMethodMethodLinksMMLU hi AccuracyFXT λ3mode=Zero-shotmode=Zero-shot2025.0724.8BPEmode=Zero-shotmode=Zero-shot2025.0724.3BINOMIALmode=Zero-shotmode=Zero-shot2025.0724.3