ResearchTasksAttack and collusion detection catch rate monitoringAttack and collusion detection catch rate monitoringFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedBashArena (held-out 80% honest rollouts)Bootstrapped Monitoring69T-only Catch Rate24Jun 11, 2026