ResearchDatasetsR-judgeR-judgeFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsAgent SafetyR-Judge97.3Accuracy92Trajectory-level safety evaluationR-judge (test)95.2Accuracy32Binary safe/unsafe classificationR-Judge (test)57.8Accuracy4