AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

Expected Returns and Policy Inconsistency-Aware Offline Federated Deep Reinforcement Learning

Meng XU, Zhongying Chen, Weiwei Fu, Yan Li, Shuguang Wang, Jianping Wang

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning

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 policy-inconsistency and Q-value aggregation plus decay to existing offline FDRL methodsFDQL · D4RL HalfCheetahfinal D4RL score: 42.78 -> 48.55Improved
add the framework to FEDORAFEDORA · D4RL Walker2dfinal D4RL score: 67.99 -> 109.29Improved
add the framework to CSCCSC · D4RL Walker2dfinal D4RL score: 108.56 -> 65.12Hurt
add the framework to FOVAFOVA · D4RL HalfCheetahfinal D4RL score: 46.99 -> 42.35Hurt
add the framework to FDRLDRFDRLDR · D4RL Hopperfinal D4RL score: 44.44 -> 43.11Hurt
weight clients by both Q-value and policy inconsistency instead of policy inconsistency only (GAPI) (ablation)Ours+FDRLDR/FORL · D4RL HalfCheetahfinal D4RL score: 50.27 -> 51.74 (FORL)Improved
weight clients by both criteria instead of Q-value only (GAQ) (ablation)Ours+FORL · D4RL HalfCheetahfinal D4RL score: 43.4 -> 51.74Improved
add decay factor on global model influence (zeta=0.99) versus none (ablation)Ours+FORL · D4RL HalfCheetahfinal D4RL score: 48.02 -> 51.74Improved
use stronger decay zeta 0.8 or 0.9 instead of 0.99 (ablation)Ours+FORL · D4RL HalfCheetahfinal D4RL score: 51.74 -> 46.84 (0.8), 44.34 (0.9)Hurt
use JSD instead of squared difference for policy inconsistency (ablation)Ours+FORL · D4RL HalfCheetahfinal D4RL score: 49.2 -> 51.74Improved
use JSD instead of KLD for policy inconsistency (ablation)Ours+FORL · D4RL HalfCheetahfinal D4RL score: 49.76 -> 51.74Improved
use multiplicative importance (scaled Q times 1/Dis) instead of subtractive (ablation)Ours+FORL · D4RL HalfCheetahfinal D4RL score: 51.74 -> 2.66Hurt

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