ResearchBenchmarksSafe Reinforcement Learning on DSRL MediumMeanSafe Reinforcement Learning on DSRL MediumMeanFollow236.57RewardCDCP49.962898.4089146.855195.3011Jul 4, 2026Evaluation resultsMethodMethodLinksRewardCostCDCP2026.07236.5719.65CMTDiff2026.07213.3922.2CMTDD2026.0757.1420.71