AI Paper Atlas

Research map / Reinforcement learning / Offline RL

DR-SAC: Distributionally Robust Soft Actor-Critic for Reinforcement Learning under Uncertainty

Mingxuan Cui, Duo Zhou, Yuxuan Han, Grani A. Hanasusanto, Qiong Wang, Huan Zhang, Zhengyuan Zhou

ICLR 2026 · Reinforcement learning › Offline RL · approach: Offline RL sample complexity theory · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add distributionally robust Bellman updateSAC · Pendulum (length +-20%)average reward: 35%Improved
add distributionally robust Bellman updateSAC · LunarLander (20% perturbation)average reward: 240 vs under 180Improved
add distributionally robust Bellman updateSAC · LunarLander (parameters -30%)reward: 9.8xImproved
add distributionally robust Bellman updateSAC · HalfCheetah (back_damping within 50%)average reward: 5950 -> 6300Improved
use functional approximation instead of accurate Bellman operator (ablation)DR-SAC · robustness environmentsrobustness and training time: less than 2% training timeImproved
use DR-SAC instead of RFQIoffline DR-RL · training timetraining time: 93 -> 4 min (Cartpole)Improved
use 5 GD steps instead of RFQI's 1000 (ablation)RFQI-style update · training efficiencyperformance: 1000 -> 5 GD stepsImproved
vary VAE latent dimension between 5 and 20 (ablation)DR-SAC · PendulumrobustnessNo clear effect
replace VAE with normalizing flow (ablation)DR-SAC · PendulumrobustnessHurt
replace VAE with diffusion model (ablation)DR-SAC · Pendulumrobustness / training time: at least 4.5x training timeNo clear effect
reduce RFQI gradient descent steps (ablation)RFQI · unperturbed environmentsperformanceHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy