ResearchBenchmarksReinforcement Learning on MuJoCo Hopper epsilon=0.075 (test)Reinforcement Learning on MuJoCo Hopper epsilon=0.075 (test)Follow3,684Natural ReturnBAR3,056.883,219.693,382.53,545.31Dec 1, 2025Evaluation resultsMethodMethodLinksNatural ReturnRandom Attack ReturnCritic Attack ReturnMAD Attack ReturnRS Attack ReturnSA-RL Attack ReturnPA-AD Attack ReturnWorst ReturnRobustness ScoreBAR2025.123,6843,6923,6783,4371,3401,7281,9081,340-0.636WocaR2025.123,6293,6373,6573,1501,1711,4522,1241,171-0.677SA2025.123,5182,8353,6623,0451,4071,4761,2861,286-0.634RADIAL2025.123,2543,1703,7062,5581,3079931,696993-0.718PPO2025.123,0812,9232,0351,7637567982379-0.974