Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Maze2D

Benchmarks

Task NameDataset NameSOTA ResultTrend
Offline Reinforcement LearningMaze2D medium
Normalized Return179.2
38
Offline Reinforcement LearningMaze2D umaze
Normalized Return141
38
Offline Reinforcement LearningMaze2D large
Normalized Return96.8
33
Offline Reinforcement LearningMaze2D large v1
Normalized Return220.66
30
Offline Reinforcement LearningMaze2D medium v1
Normalized Return166.82
30
State ExplorationMaze2D Square-b
State Coverage Ratio85
22
Robotic Path PlanningMaze2D (test)
BS1-1
22
Offline Reinforcement LearningMaze2D umaze v1
Normalized Return52.2
18
Planning and Controlmaze2d-umaze v1 (100 episodes, 300 steps/ep)
Score165.19
16
Reward Conditioning (RC)Maze2D (test)
Reward2.74
16
Behavior Cloning (BC)Maze2D (test)
Reward2.74
16
State ExplorationMaze2D Square-tree
State Coverage Ratio50
11
State ExplorationMaze2D Corridor2
State Coverage Ratio93
11
State ExplorationMaze2D Square-d
State Coverage Ratio0.77
11
State ExplorationMaze2D Square-c
State Coverage Ratio74
11
State ExplorationMaze2D Square-a
State Coverage Ratio87
11
Long horizon planningMaze2D U-Maze
Normalized Return185.3
10
Offline Reinforcement LearningMaze2D large v0 (test)
Score187.8
10
Offline Reinforcement LearningMaze2D medium v0 (test)
Score152.3
10
Offline Reinforcement LearningMaze2D umaze v0 (test)
Overall Score111
10
Continuous ControlMaze2D large
Total Reward361
9
Continuous ControlMaze2D medium
Total Reward416.28
9
Continuous ControlMaze2D umaze
Total Reward182.1
9
NavigationMaze2d (test)
Average Success Rate87.31
8
Goal-Conditioned Trajectory PlanningMaze2D U-Maze
Success Score122.3
8
Showing 25 of 50 rows