ResearchBenchmarksRole-playing Quality Evaluation on RoleChat (Human Evaluation)Follow87Win RateRoleJudge78.6880.848385.16Apr 15, 2026Evaluation ResultsMethodMethodLinksWin RateLose RateRoleJudgeComparison Baseline=Qw...Comparison Baseline=Qwen3-Omni2026.048713RoleJudgeComparison Baseline=Ge...Comparison Baseline=Gemini3 Pro2026.047921