ResearchBenchmarksClassification on DTFollow72AccuracyCARO50.1655.8361.567.17Feb 28, 2026Evaluation ResultsMethodMethodLinksAccuracyCohen's KappaCAROPrompting Strategy=Con...Prompting Strategy=Confusion-Aware Rubric Optimization2026.02720.53GradeOptPrompting Strategy=Gra...Prompting Strategy=GradeOpt aggregate feedback2026.02650.44NaivePrompting Strategy=Nai...Prompting Strategy=Naive prompting2026.02510.27