ResearchTasksSafeguarding LLM Agents against prompt injectionSafeguarding LLM Agents against prompt injectionFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedBanking and Slack (test)Repeated-Prompt91.9BU (No Attack)21May 6, 2026