PointPush1 on Safety-Gymnasium
49Normalized RewardFISOR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FISOR2026.04 | 49 | 83 | |
| BCQ-Lag2026.04 | 33 | 86 | |
| BCQ-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 33 | 86 | |
| DT + SASBackbone=DT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 28 | 62 | |
| DT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 28 | 62 | |
| CDT+SASAlignment=SAS, Base Model=CDT2026.04 | 26 | 54 | |
| CDT + SASBackbone=CDT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 26 | 54 | |
| CDT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 26 | 54 | |
| CDTBackbone=CDT2026.04 | 24 | 48 | |
| CDTcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 24 | 48 | |
| BEAR-Lag2026.04 | 22 | 79 | |
| BEAR-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 22 | 79 | |
| CPQ2026.04 | 20 | 83 | |
| CPQcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 20 | 83 | |
| BC-All2026.04 | 19 | 61 | |
| BC-Allcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 19 | 61 | |
| BC-Safe2026.04 | 13 | 43 | |
| COptiDICE2026.04 | 13 | 83 | |
| BC-Safecost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 13 | 43 | |
| COptiDICEcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 13 | 83 | |
| DCRL2026.04 | 1 | 52 |