ResearchBenchmarksAgent Believability on SOTOPIA AURATownFollow3.75Mean Human RatingAURA2.09642.52572.9553.3843Jun 4, 2026Evaluation ResultsMethodMethodLinksMean Human RatingMean Difference (Rater Means)Wilcoxon p-value (n=8)95% Cluster CI Lower BoundAURARating Scale=5-point L...Rating Scale=5-point Likert, Raters=8, Scenarios=50, Warmup=10 steps, Seed=422026.063.751.590.0170.88VanillaRating Scale=5-point L...Rating Scale=5-point Likert, Raters=8, Scenarios=50, Warmup=10 steps, Seed=422026.062.16———