Safe Continual Reinforcement Learning on Damaged Ant Velocity benchmark leaderboard · SOTA2 Research

Safe Continual Reinforcement Learning on Damaged Ant Velocity

3,739.7Final RewardCF-EWC

Evaluation Results

Method	Links
CF-EWC2026.04		3,739.7	0.6	11,208.5
PPO+EWC2026.04		3,645.3	0.8	11,061.7
PPO2026.04		3,282.1	2.2	9,181
CPO2026.04		2,708.7	30.3	886.4
Safe EWC2026.04		2,689.9	5.1	312.4
CPPO-PID2026.04		2,688.8	1.3	800.6
PPO-Lag2026.04		2,676.4	0.1	1,009.7