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

Safe Continual Reinforcement Learning on Damaged HalfCheetah Velocity

2,078.8Final RewardPPO

Evaluation Results

Method	Links
PPO2026.04		2,078.8	0.242	7,368.4
CF-EWC2026.04		2,067.1	0.221	6,293.5
PPO+EWC2026.04		2,036.6	0.242	6,422.6
Safe EWC2026.04		1,631.3	0.18	393.1
PPO-Lag2026.04		1,553.5	0.284	830.6
CPO2026.04		1,536.3	0.588	495.5
CPPO-PID2026.04		1,487.8	0.242	576.5
Exp. Replay2026.04		504.9	0.173	666.7