| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Scenario Generation | MetaDrive | Attack Rate21.1 | 16 | |
| Safe Reinforcement Learning | MetaDrive | Normalized Reward0.69 | 10 | |
| Autonomous Driving | MetaDrive-Keyboard (test) | Episodic Return347.523 | 6 | |
| Autonomous Driving | MetaDrive Keyboard (train) | Total Safety Cost30,000 | 6 | |
| Offline Safe Reinforcement Learning | MetaDrive | Normalized Reward0.74 | 5 | |
| Goal Navigation | MetaDrive 4-Goal | Success Rate47.1 | 5 | |
| Safe Reinforcement Learning | MetaDrive Hard (held-out seeds 10000–10019) | Categorical Violation Rate (%)25 | 4 | |
| Autonomous Driving | MetaDrive DSRL | Reward0.62 | 4 | |
| Decision Making | MetaDrive (test) | Normalized Score1.02 | 4 | |
| Safe Reinforcement Learning | MetaDrive Medium (held-out seeds 10000–10019) | Categorical Success Rate41 | 2 | |
| Safe Reinforcement Learning | MetaDrive Easy (held-out seeds 10000–10019) | Categorical Violation Rate (%)30 | 2 |