ResearchDatasetsR-judgeFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsAgent SafetyR-Judge97.3Accuracy92Trajectory-level safety evaluationR-judge (test)95.2Accuracy32Binary safe/unsafe classificationR-Judge (test)57.8Accuracy4