ResearchTasksHarmlessness preference labeling accuracyHarmlessness preference labeling accuracyFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedSafeRLHF-RMB (test)Biased Rubric Search70.6Bench Accuracy15Feb 26, 2026