# Safe Langevin Soft Actor Critic

arXiv: https://arxiv.org/abs/2602.00587 · HTML: https://mlsanity.com/paper/2602.00587/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in RL theory and sample complexity, one of 15 directions in Deep reinforcement learning methods.
- Builds on 18 of its 33 references that are in the map.
- First posted on arXiv 31 Jan 2026.

## Placed in

- Deep reinforcement learning methods / [RL theory and sample complexity](https://mlsanity.com/deeprl/rl-theory-and-sample-complexity/)

## Builds on (most cited first, 18 of 18)

- [Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor](https://mlsanity.com/paper/1801.01290/) (2018)
- [Playing Atari with Deep Reinforcement Learning](https://mlsanity.com/paper/1312.5602/) (2013)
- [Addressing Function Approximation Error in Actor-Critic Methods](https://mlsanity.com/paper/1802.09477/) (2018)
- [Constrained Policy Optimization](https://mlsanity.com/paper/1705.10528/) (2017)
- [End-to-End Training of Deep Visuomotor Policies](https://mlsanity.com/paper/1504.00702/) (2015)
- [A Distributional Perspective on Reinforcement Learning](https://mlsanity.com/paper/1707.06887/) (2017)
- [QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation](https://mlsanity.com/paper/1806.10293/) (2018)
- [Implicit Quantile Networks for Distributional Reinforcement Learning](https://mlsanity.com/paper/1806.06923/) (2018)
- [Risk-Constrained Reinforcement Learning with Percentile Risk Criteria](https://mlsanity.com/paper/1512.01629/) (2015)
- [Responsive Safety in Reinforcement Learning by PID Lagrangian Methods](https://mlsanity.com/paper/2007.03964/) (2020)
- [MetaDrive: Composing Diverse Driving Scenarios for Generalizable Reinforcement Learning](https://mlsanity.com/paper/2109.12674/) (2021)
- [Identifying and attacking the saddle point problem in high-dimensional non-convex optimization](https://mlsanity.com/paper/1406.2572/) (2014)
- [Projection-Based Constrained Policy Optimization](https://mlsanity.com/paper/2010.03152/) (2020)
- [Safety-Gymnasium: A Unified Safe Reinforcement Learning Benchmark](https://mlsanity.com/paper/2310.12567/) (2023)
- [Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks](https://mlsanity.com/paper/1512.07666/) (2015)
- [OmniSafe: An Infrastructure for Accelerating Safe Reinforcement Learning Research](https://mlsanity.com/paper/2305.09304/) (2023)
- [Worst Cases Policy Gradients](https://mlsanity.com/paper/1911.03618/) (2019)
- [Langevin Soft Actor-Critic: Efficient Exploration through Uncertainty-Driven Critic Learning](https://mlsanity.com/paper/2501.17827/) (2025)

## Built on by (most cited first, 1 of 1)

- [Safe on Average, Unsafe in the Tail: When Is the Episodic-Cost Tail Controllable?](https://mlsanity.com/paper/2610.09508/) (2026)

## Abstract (arXiv)

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.