CarPush2 on Safety Gymnasium
0.24RewardCPQ
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CPQ2026.04 | 0.24 | 4.25 | |
| CPQcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.24 | 4.25 | |
| DT + SASBackbone=DT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 0.22 | 1.16 | |
| DT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.22 | 1.16 | |
| CDT + SASBackbone=CDT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 0.21 | 0.75 | |
| CDT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.21 | 0.75 | |
| CDTBackbone=CDT2026.04 | 0.19 | 1.3 | |
| CDTcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.19 | 1.3 | |
| BCQ-Lag2026.04 | 0.15 | 1.38 | |
| BCQ-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.15 | 1.38 | |
| BC-All2026.04 | 0.14 | 0.9 | |
| BC-Allcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.14 | 0.9 | |
| BEAR-Lag2026.04 | 0.1 | 1.2 | |
| BEAR-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.1 | 1.2 | |
| COptiDICE2026.04 | 0.09 | 1.07 | |
| COptiDICEcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.09 | 1.07 | |
| BC-Safe2026.04 | 0.05 | 0.45 | |
| BC-Safecost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 0.05 | 0.45 | |
| DCRL2026.04 | -0.13 | 0.17 |