AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Q-SAM: Unlocking Sharpness-Aware Minimization for Generalization in Offline Reinforcement Learning

Da Wang, Yi Ma, Ting Guo, Lin Li, Wei Wei, Jiye Liang

ICML 2026 · Reinforcement learning › Offline RL · approach: Offline RL sample complexity theory

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
replace Adam with plain SAM for the policyTD3BC · D4RL gym MuJoCo-v2 totalTotal normalized score: 688.0 -> 693.8No clear effect
replace Adam with Q-SAM for the policyTD3BC · D4RL AntMaze-v2 totalTotal normalized score: 116.2 -> 163.4Improved
replace Adam with Q-SAM for the policyIQL · D4RL gym MuJoCo-v2 totalTotal normalized score: 734.6 -> 772.1Improved
replace Adam with Q-SAM for the policyCQL · D4RL AntMaze-v2 totalTotal normalized score: 304.5 -> 335.0Improved
add weighted sharpness (lambda-weighted SAM) (ablation)Q-SAM on CQL · D4RL gym MuJoCo-v2 totalTotal normalized score: 730.8 -> 745.0Improved
add Q-bound weighted samples (ablation)Q-SAM on CQL · D4RL gym MuJoCo-v2 totalTotal normalized score: 736.5 -> 745.0Improved
add Q-bound weighted samples on halfcheetah datasets (ablation)Q-SAM on CQL · halfcheetah-mr / halfcheetah-menormalized score: drop -0.1 and -0.3 without weighted samplesNo clear effect
use SAM instead of Adam for the policyTD3BC · hopper-medium-replayLocal Action Sensitivity: 1.36 -> 0.67Improved
double TD3BC training steps from 1M to 2M instead of using Q-SAM (ablation)TD3BC · hopper-medium-replaynormalized score: 64.4 -> 65.4 (TD3BC + Q-SAM at 1M: 87.3)No clear effect
apply Q-SAM with small-sample data (5% of dataset)TD3BC · halfcheetah-medium-replay 5% datanormalized scoreImproved
add weighted sharpness to reduce Bellman error (ablation)Q-SAM · critic loss (small-sample)Log(1 + Critic Loss)Improved
Q-SAM on TD3BC for hard antmaze tasksTD3BC · antmaze-large-play and large-diversenormalized score: 0.0 -> 0.0No clear effect

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