ResearchTasksDeceptive Alignment / Surveillance RobustnessDeceptive Alignment / Surveillance RobustnessFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedAI Safety Gridworlds Absent Supervisor (final-round)EPO-Safe41R Score8Apr 28, 2026