ResearchBenchmarksSafe Reinforcement Learning on Safety Gymnasium Goal (threshold=20)Follow24.26ReturnPPO-BT5.040810.030415.0220.0096Mar 24, 2026Evaluation ResultsMethodMethodLinksReturnCostPPO-BT2026.0324.2626.97PPO-LagType=OracleType=Oracle2026.0323.520.7PbCRL2026.0323.4118.5RLSF2026.035.788.2