Behavior Cloning on D4RL AntMaze navigation tasks v2
84.2AntMaze U Success RateGTP-BC
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GTP-BCevaluation episodes=100, sampling steps (K)=5, value-guidance coefficient (η)=02025.10 | 84.2 | 79.2 | 74.4 | 85 | 34.4 | 40.8 | 66.3 | |
| Diffuserevaluation episodes=100, sampling steps (K)=52025.10 | 78.9 | 55 | 0 | 0 | 6.7 | 2.2 | 23.8 | |
| TD3+BCevaluation episodes=1002025.10 | 78.6 | 71.4 | 10.6 | 3 | 0.2 | 0 | 27.3 | |
| C-BCevaluation episodes=100, sampling steps (K)=22025.10 | 75.8 | 77.6 | 56.8 | 31.6 | 10.2 | 12.8 | 44.1 | |
| MoRelevaluation episodes=1002025.10 | 73 | 61 | 0 | 8 | 0 | 0 | 23.7 | |
| D-BCevaluation episodes=100, sampling steps (K)=52025.10 | 71.8 | 61.2 | 43.4 | 29.8 | 14.6 | 26.6 | 41.2 | |
| Onestep RLevaluation episodes=1002025.10 | 64.3 | 60.7 | 0.3 | 0 | 0 | 0 | 20.9 | |
| DTevaluation episodes=1002025.10 | 59.2 | 53 | 0 | 0 | 0 | 0 | 18.7 | |
| AWACevaluation episodes=1002025.10 | 56.7 | 49.3 | 0 | 0.7 | 0 | 1 | 18 | |
| BCevaluation episodes=1002025.10 | 54.6 | 45.6 | 0 | 0 | 0 | 0 | 16.7 |