ResearchBenchmarksReinforcement Learning on Walker2D MuJoCo (unseen states)Reinforcement Learning on Walker2D MuJoCo (unseen states)Follow3,200RewardGATN2,1602,4302,7002,970Jul 2, 2025Evaluation resultsMethodMethodLinksRewardGeneralization GapRobustnessGATN2025.073,20020085A2T2025.072,80030075PNN2025.072,60035070DQN2025.072,20050060