ResearchBenchmarksSafe Reinforcement Learning on SafeAntMaze WshapeFollow35Final Success RateITES-1.48.0517.526.95Jun 21, 2026Evaluation ResultsMethodMethodLinksFinal Success RateFinal CostITES2026.0635183TD3L/R2026.065154CUP2026.060103.5FOCOPS2026.060141TD3LAG2026.060128.68PPOLAG2026.060105SafeDreamer2026.060216.3