ResearchBenchmarksInstruction Following on Muses-Bench Conflict n=368Follow86.98AccuracyDeepseek28.604843.759958.91574.0701Jun 20, 2026Evaluation ResultsMethodMethodLinksAccuracyWeighted AccuracyPer-User AccuracyDeepseekFramework=Muses-Bench,...Framework=Muses-Bench, Mode=Conflict2026.0686.9887.2215.57QwenFramework=Muses-Bench,...Framework=Muses-Bench, Mode=Conflict2026.0685.0585.0515.21GeminiFramework=Muses-Bench,...Framework=Muses-Bench, Mode=Conflict2026.0658.1166.6111.73GPTFramework=Muses-Bench,...Framework=Muses-Bench, Mode=Conflict2026.0630.8543.847.92