Fernando Acero

Entropy Regularization Improves Policy Robustness in Continuous-Time Reinforcement Learning

Jialun Cao, Fernando Acero, David Šiška, Yufei Zhang

Entropy regularization is widely used in continuous-time reinforcement learning (RL) to reduce sensitivity to environmental perturbations, yet its robustness benefits lack a rigorous theoretical foundation. This paper establishes the first robustness guarantees for entropy-regula…Read more