Behavior Cloning on D4RL Gym locomotion tasks v2
48.6Return (HalfCheetah-m)GTP-BC
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| GTP-BCevaluation episodes=10, sampling steps (K)=5, value-guidance coefficient (η)=02025.10 | 48.6 | 83.7 | 77.1 | 46.3 | 100.5 | 83.4 | 91.3 | 109.6 | 100.2 | 82.3 | |
| Onestep RLevaluation episodes=102025.10 | 48.4 | 59.6 | 81.8 | 38.1 | 97.5 | 49.5 | 93.4 | 103.3 | 113 | 76.1 | |
| TD3+BCevaluation episodes=102025.10 | 48.3 | 59.3 | 83.7 | 44.6 | 60.9 | 81.8 | 90.7 | 98 | 110.1 | 75.3 | |
| D-BCevaluation episodes=10, sampling steps (K)=52025.10 | 45.4 | 65.3 | 81.2 | 41.7 | 67.3 | 77.5 | 90.8 | 107.6 | 108.9 | 76.3 | |
| Diffuserevaluation episodes=10, sampling steps (K)=52025.10 | 44.2 | 58.5 | 79.7 | 42.2 | 96.8 | 61.2 | 79.8 | 107.2 | 108.4 | 75.3 | |
| AWACevaluation episodes=102025.10 | 43.5 | 57 | 72.4 | 40.5 | 37.2 | 27 | 42.8 | 55.8 | 74.5 | 50.1 | |
| BCevaluation episodes=102025.10 | 42.6 | 52.9 | 75.3 | 36.6 | 18.1 | 26 | 55.2 | 52.5 | 107.5 | 51.9 | |
| DTevaluation episodes=102025.10 | 42.6 | 67.6 | 74 | 36.6 | 82.7 | 66.6 | 86.8 | 107.6 | 108.1 | 74.7 | |
| MoRelevaluation episodes=102025.10 | 42.1 | 95.4 | 77.8 | 40.2 | 93.6 | 49.8 | 53.3 | 108.7 | 95.6 | 72.9 | |
| C-BCevaluation episodes=10, sampling steps (K)=22025.10 | 31 | 71.7 | 83.1 | 34.4 | 99.7 | 73.3 | 32.7 | 90.6 | 110.4 | 69.7 |