ResearchTasksAttack PreventionAttack PreventionFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedAGENTREDBENCH cross-integration held-out (test)AGENTREDGUARD-MiniLM32.1Baseline ASR9Jun 2, 2026