ResearchBenchmarksOffline Reinforcement Learning on MuJoCo and Maze2D Step-wise Reward v2Offline Reinforcement Learning on MuJoCo and Maze2D Step-wise Reward v2Follow66.5Hopper-Medium ScoreQDT57.6659.95562.2564.545Dec 25, 2025Evaluation resultsMethodMethodLinksHopper-Medium ScoreHopper-Medium-Replay ScoreWalker2d-Medium ScoreWalker2d-Medium-Replay ScoreHalfCheetah-Medium ScoreHalfCheetah-Medium-Replay ScoreMaze2D-UMaze ScoreMaze2D-Medium ScoreMaze2D-Large ScoreQDTReward Type=Step-wise...Reward Type=Step-wise Reward2025.1266.552.167.158.242.335.657.313.331IQLReward Type=Step-wise...Reward Type=Step-wise Reward2025.1263.892.179.973.747.444.137.735.549.6DTReward Type=Step-wise...Reward Type=Step-wise Reward2025.1260.363.773.360.242.134.1318.22.3CQLReward Type=Step-wise...Reward Type=Step-wise Reward2025.125848.679.274.144.446.25.7512.5