ResearchBenchmarksToxicity Classification on DICES 3-class (item-level)Follow67.79Accuracy (Acc)DiADEM65.91866.40466.8967.376Apr 9, 2026Evaluation ResultsMethodMethodLinksAccuracy (Acc)F1 Score (Macro)F1 Score (Weighted)Cohen's KappaMCCJSDMDError Rate (ER)ECEDiADEM2026.0467.7938.0262.8317.2619.310.022982.6658.954.84DisCo2026.0466.2837.3761.7814.8616.20.026684.7161.984.7LeWiDi2026.0465.9938.8762.7718.2118.870.023388.362.582.14