Offline Reinforcement Learning on Maze2D large v1
220.66Normalized ReturnInverter
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| InverterK=16, #NN-passes=502026.05 | 220.66 | 0.2 | 1.59 | 14.3 | 93.7 | 1 | |
| AWAC#NN-passes=8002026.05 | 209.13 | 8.19 | 1.28 | — | 1.02 | 10.9 | |
| SAC-N#NN-passes=8002026.05 | 204.76 | 1.19 | 1.8 | — | 1.44 | 15.4 | |
| Diffuser#NN-passes=2562026.05 | 123.07 | 62.22 | 14.32 | 17 | 3.68 | 39.3 | |
| TD3+BC#NN-passes=8002026.05 | 97.1 | 25.41 | 1.67 | — | 1.34 | 14.3 | |
| EDAC#NN-passes=8002026.05 | 95.6 | 22.92 | 1.74 | — | 1.39 | 14.9 | |
| ReBRAC#NN-passes=8002026.05 | 78.33 | 61.77 | 0.21 | — | 168 | 1.8 | |
| IQL#NN-passes=8002026.05 | 61.72 | 3.5 | 1.69 | — | 1.36 | 14.5 | |
| PrivORL-jRL Algorithm=IQL, epsilon=102025.12 | 37.7 | — | — | — | — | — | |
| DT#NN-passes=8002026.05 | 35.66 | 28.2 | 3.57 | — | 2.86 | 30.5 | |
| PrivORL-jRL Algorithm=TD3PLUSBC, epsilon=102025.12 | 35.6 | — | — | — | — | — | |
| PrivORL-j-URL Algorithm=IQL, epsilon=102025.12 | 31.9 | — | — | — | — | — | |
| PrivORL-jRL Algorithm=EDAC, epsilon=102025.12 | 30.5 | — | — | — | — | — | |
| PrivORL-jRL Algorithm=TD3PLUSBC, epsilon=12025.12 | 29.5 | — | — | — | — | — | |
| DP-TransformerRL Algorithm=TD3PLUSBC, epsilon=102025.12 | 28.3 | — | — | — | — | — | |
| CQL#NN-passes=8002026.05 | 23.75 | 36.7 | 1.84 | — | 1.47 | 15.7 | |
| DP-TransformerRL Algorithm=IQL, epsilon=102025.12 | 23.4 | — | — | — | — | — | |
| DP-TransformerRL Algorithm=EDAC, epsilon=102025.12 | 21.8 | — | — | — | — | — | |
| PrivORL-jRL Algorithm=EDAC, epsilon=12025.12 | 20.5 | — | — | — | — | — | |
| PrivORL-j-URL Algorithm=EDAC, epsilon=102025.12 | 19.8 | — | — | — | — | — | |
| PrivORL-j-URL Algorithm=TD3PLUSBC, epsilon=102025.12 | 14.3 | — | — | — | — | — | |
| BC-10%#NN-passes=8002026.05 | 11.32 | 5.1 | 1.25 | — | 999.7 | 10.7 | |
| PrivORL-j-URL Algorithm=EDAC, epsilon=12025.12 | 10.9 | — | — | — | — | — | |
| PrivORL-jRL Algorithm=IQL, epsilon=12025.12 | 9.9 | — | — | — | — | — | |
| PrivORL-j-URL Algorithm=TD3PLUSBC, epsilon=12025.12 | 9.9 | — | — | — | — | — | |
| PrivORL-j-URL Algorithm=IQL, epsilon=12025.12 | 9.8 | — | — | — | — | — | |
| DP-TransformerRL Algorithm=EDAC, epsilon=12025.12 | 7.6 | — | — | — | — | — | |
| DP-TransformerRL Algorithm=TD3PLUSBC, epsilon=12025.12 | 6.3 | — | — | — | — | — | |
| DP-TransformerRL Algorithm=IQL, epsilon=12025.12 | 4.9 | — | — | — | — | — | |
| BC#NN-passes=8002026.05 | 2.26 | 4.39 | 1.7 | — | 1.36 | 14.6 |