ResearchBenchmarksChat Dialogue Evaluation on ChatCOREFollow36.1ScoreUnigram35.26835.48435.735.916May 21, 2026Evaluation ResultsMethodMethodLinksScoreUnigramEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0536.1BPEEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0535.78WordPieceEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0535.77ToaSTEvaluation Protocol=Su...Evaluation Protocol=Supervised fine-tuning2026.0535.3