Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Safe Reinforcement Learning on MetaDrive Hard (held-out seeds 10000–10019)
Loading...
25
Categorical Violation Rate (%)
VLMPPOLag+Conf
24.68
26.84
29
31.16
Jun 9, 2026
Categorical Violation Rate (%)
Total Violation Rate (%)
Categorical Violation Rate Difference (pp)
Significance Test Result
Updated 1mo ago
Evaluation Results
Method
Method
Links
Categorical Violation Rate (%)
Total Violation Rate (%)
Categorical Violation Rate Difference (pp)
Significance Test Result
VLMPPOLag+Conf
warm-start λ0=0.5
2026.06
25
28
26
-
VLMPPOLag+Conf
warm-start λ0=default
2026.06
31
39
-
-
PPOLag
warm-start λ0=default
2026.06
33
36
-
-
PPOLag
warm-start λ0=0.5
2026.06
33
36
-
-
Feedback
Search any
task
Search any
task