ResearchTasksLLM Agent Security EvaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedAgentDojo vulnerable-agent profile stress test (six banking pairs)Deny0Attack Success Count (out of 6)5May 15, 2026Agent Security Bench (test)Repeat prompt73.67Benign Utility (BU)5Mar 24, 2026
AgentDojo vulnerable-agent profile stress test (six banking pairs)Deny0Attack Success Count (out of 6)5May 15, 2026