ResearchBenchmarksMultitask Language Understanding on MMLU (Aggregate Score)Follow36.35MMLU Aggregate ScoreUnigram36.006836.095936.18536.2741May 21, 2026Evaluation ResultsMethodMethodLinksMMLU Aggregate ScoreUnigramEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0536.35ToaSTEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0536.13WordPieceEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0536.09BPEEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0536.02