Safe Reinforcement Learning on MetaDrive Easy (held-out seeds 10000–10019)
30Categorical Violation Rate (%)PPOLag
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PPOLag2026.06 | 30 | 43 | — | — | |
| VLMPPOLag+Conf2026.06 | 35 | 38 | — | 3 |
| Method | Links | ||||
|---|---|---|---|---|---|
| PPOLag2026.06 | 30 | 43 | — | — | |
| VLMPPOLag+Conf2026.06 | 35 | 38 | — | 3 |