ResearchTasksOffline Constrained RLHFOffline Constrained RLHFFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedPKU-SafeRLHF 74k (train)πλ-1.603Safety Expected Reward (r1)2Apr 2, 2026PKU-SafeRLHF 7k (test)πλ-1.604Safety Reward (r1)1Apr 2, 2026