ResearchDatasetsAlpaca + XSTestAlpaca + XSTestFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsSafety DetectionAlpaca + XSTest full (val-selected)0.991AUROC42Refusal DetectionAlpaca + XSTest (val)0.988AUROC42