English

Safe Langevin Soft Actor Critic

Machine Learning 2026-02-03 v1

Abstract

Balancing reward and safety in constrained reinforcement learning remains challenging due to poor generalization from sharp value minima and inadequate handling of heavy-tailed risk distribution. We introduce Safe Langevin Soft Actor-Critic (SL-SAC), a principled algorithm that addresses both issues through parameter-space exploration and distributional risk control. Our approach combines three key mechanisms: (1) Adaptive Stochastic Gradient Langevin Dynamics (aSGLD) for reward critics, promoting ensemble diversity and escape from poor optima; (2) distributional cost estimation via Implicit Quantile Networks (IQN) with Conditional Value-at-Risk (CVaR) optimization for tail-risk mitigation; and (3) a reactive Lagrangian relaxation scheme that adapts constraint enforcement based on the empirical CVaR of episodic costs. We provide theoretical guarantees on CVaR estimation error and demonstrate that CVaR-based Lagrange updates yield stronger constraint violation signals than expected-cost updates. On Safety-Gymnasium benchmarks, SL-SAC achieves the lowest cost in 7 out of 10 tasks while maintaining competitive returns, with cost reductions of 19-63% in velocity tasks compared to state-of-the-art baselines.

Cite

@article{arxiv.2602.00587,
  title  = {Safe Langevin Soft Actor Critic},
  author = {Mahesh Keswani and Samyak Jain and Raunak P. Bhattacharyya},
  journal= {arXiv preprint arXiv:2602.00587},
  year   = {2026}
}

Comments

20 pages, 12 figures

R2 v1 2026-07-01T09:29:11.892Z