PointPush2 on Safety-Gymnasium
44Normalized RewardFISOR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FISOR2026.04 | 44 | 11 | |
| DT + SASBackbone=DT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 24 | 0.64 | |
| DT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 24 | 0.64 | |
| BCQ-Lag2026.04 | 23 | 0.99 | |
| BCQ-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 23 | 0.99 | |
| CDTBackbone=CDT2026.04 | 21 | 0.65 | |
| CDTcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 21 | 0.65 | |
| CDT+SASAlignment=SAS, Base Model=CDT2026.04 | 20 | 53 | |
| CDT + SASBackbone=CDT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 20 | 0.53 | |
| CDT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 20 | 0.53 | |
| BC-All2026.04 | 18 | 0.91 | |
| BC-Allcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 18 | 0.91 | |
| BEAR-Lag2026.04 | 16 | 0.89 | |
| BEAR-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 16 | 0.89 | |
| BC-Safe2026.04 | 11 | 0.8 | |
| CPQ2026.04 | 11 | 1.04 | |
| BC-Safecost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 11 | 0.8 | |
| CPQcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 11 | 1.04 | |
| COptiDICE2026.04 | 2 | 1.18 | |
| DCRL2026.04 | 2 | 0.07 | |
| COptiDICEcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 2 | 1.18 |