ResearchBenchmarksSafe Reinforcement Learning on DSRL EasySparseFollow330.39RewardCMTDiff63.63132.885202.14271.395Jul 4, 2026Evaluation ResultsMethodMethodLinksRewardCostCMTDiff2026.07330.3945.52CDCP2026.07326.0625.21CMTDD2026.0773.8924.08