ResearchTasksStateful Agent Backdoor Attack EvaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedNote-based instantiation 20 episodesLlama-3.1-8B100Initial Success Rate (R_init)4May 8, 2026