ResearchBenchmarksMulti-agent interaction and social reasoning on Werewolf MultiAgentBenchFollow55.75Task PerformanceETI24.040432.272740.50548.7373Apr 21, 2026Evaluation ResultsMethodMethodLinksTask PerformanceCoordinationETIAgent=QWEN, Trait Sour...Agent=QWEN, Trait Source=QWEN2026.0455.7565.2ETIAgent=QWEN, Trait Sour...Agent=QWEN, Trait Source=GPT2026.0449.9759.52QWENTrait Source=noneTrait Source=none2026.0443.2860.2ETIAgent=GPT, Trait Sourc...Agent=GPT, Trait Source=QWEN2026.0436.4657.56ETIAgent=GPT, Trait Sourc...Agent=GPT, Trait Source=GPT2026.0429.5455.56GPTTrait Source=noneTrait Source=none2026.0425.2654.32