ResearchTasksAgent Security EvaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedAgent Security Bench (ASB) IPI Combined Attack (CA) and Context Ignoring (CI) averagedToolSafe2.89ASR18Jun 5, 2026Agent Security Bench (ASB) DPI Combined Attack (CA) and Context Ignoring (CI) averagedReAct92.42ASR18Jun 5, 2026ASB (Agent Security Benchmark)gpt-4o + GBT-SE7ASR-d (ASB)8Mar 9, 2026Agent Security Bench (ASB)No Defense84.31PNA4Apr 21, 2026Agent Security Benchmarks——Primary metric0Apr 24, 2026
Agent Security Bench (ASB) IPI Combined Attack (CA) and Context Ignoring (CI) averagedToolSafe2.89ASR18Jun 5, 2026
Agent Security Bench (ASB) DPI Combined Attack (CA) and Context Ignoring (CI) averagedReAct92.42ASR18Jun 5, 2026