Safety-constrained Reinforcement Learning on Safety-Gym SafetyPointCircle1 (evaluation)
44.3Average RewardCUP
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CUPKnowledge Level=Oracle2025.04 | 44.3 | 16.1 | — | — | — | — | |
| TraCeSKnowledge Level=Zero Knowledge2025.04 | 43.2 | 16.2 | — | — | — | 5,830 | |
| FOCOPSKnowledge Level=Oracle2025.04 | 42 | 16.3 | — | — | — | — | |
| TRPO-SauteKnowledge Level=Oracle2025.04 | 39.6 | 8.3 | — | — | — | — | |
| PPO2026.05 | 25.15 | 278.48 | 100 | 11.139 | 9 | — | |
| PPO-Lag2026.05 | 23.11 | 264.59 | 91.9 | 10.584 | 8.7 | — | |
| RiskGated2026.05 | 10.86 | 106.47 | 43.2 | 4.259 | 10.2 | — | |
| POMDP2026.05 | 9.53 | 237.23 | 37.9 | 9.489 | 4 | — | |
| FOCOPS2026.05 | 7.33 | 94.09 | 29.1 | 3.764 | 7.8 | — | |
| TRPO-Lag2026.05 | 6.62 | 107.5 | 26.3 | 4.3 | 6.2 | — | |
| RCPO2026.05 | 6.62 | 107.5 | 26.3 | 4.3 | 6.2 | — | |
| CUP2026.05 | 5.16 | 96.07 | 20.5 | 3.843 | 5.4 | — | |
| CPO2026.05 | 4.23 | 61.26 | 16.8 | 2.45 | 6.9 | — | |
| CPPO-PID2026.05 | 4.2 | 77.66 | 16.7 | 3.106 | 5.4 | — | |
| PCPO2026.05 | 3.13 | 51.65 | 12.4 | 2.066 | 6.1 | — |