PointButton2 on Safety-Gymnasium
58Normalized RewardCPQ
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CPQ2026.04 | 58 | 4.3 | |
| CPQcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 58 | 4.3 | |
| DT + SASBackbone=DT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 51 | 1.14 | |
| DT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 51 | 1.14 | |
| CDTBackbone=CDT2026.04 | 46 | 1.57 | |
| CDTcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 46 | 1.57 | |
| BEAR-Lag2026.04 | 43 | 2.47 | |
| BEAR-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 43 | 2.47 | |
| CDT+SASAlignment=SAS, Base Model=CDT2026.04 | 41 | 98 | |
| CDT + SASBackbone=CDT, Alignment strategy (SAS)=Lyapunov-Guided2026.04 | 41 | 0.98 | |
| CDT + SAScost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 41 | 0.98 | |
| BCQ-Lag2026.04 | 40 | 2.66 | |
| BCQ-Lagcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 40 | 2.66 | |
| BC-All2026.04 | 27 | 2.02 | |
| BC-Allcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 27 | 2.02 | |
| DCRL2026.04 | 18 | 0.64 | |
| BC-Safe2026.04 | 16 | 1.1 | |
| BC-Safecost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 16 | 1.1 | |
| COptiDICE2026.04 | 15 | 1.51 | |
| COptiDICEcost thresholds=3, evaluation episodes=20, random seeds=32026.04 | 15 | 1.51 | |
| FISOR2026.04 | 1 | 58 |