ResearchTasksGuarded Agent EvaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedAgentHarm latest (full)ReAct-llamafirewall97.16Refusal Rate14Feb 26, 2026ASB latest (IPI)ReAct-llamafirewall5.5ASR14Feb 26, 2026ASB latest (DPI)ReAct95.25ASR14Feb 26, 2026AgentDojo full latestReAct0.562ASR14Feb 26, 2026