AI Paper Atlas

Research map / Federated learning and privacy / Communication-efficient FL

Breaking the Capacity Bottleneck in Model-Heterogeneous Federated Learning via Gradual Model Restoration

Chengjie Ma, Seungeun Oh, Jihong Park, Seong-Lyun Kim

ICML 2026 · Federated learning and privacy › Communication-efficient FL · approach: Efficient local training in FL · 0 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
gradually restore sub-model density (GMR) instead of fixed density (ablation)FedGMR · High heterogeneity, Non-IID, four taskstest accuracy over aggregation stepsImproved
use asynchronous coordination (ablation)FedGMR · FEMNIST (High heterogeneity, Non-IID)training speedImproved
use asynchronous coordination (ablation)FedGMR · ImageNet-100 and StackOverflowaccuracyNo clear effect
apply GMR to other MHFL methods (ablation)HeteroFL/FedRolex/Fjord/Fiarse · FEMNISTaccuracyImproved
pair GMR with MaskFedAvg aggregation instead of gradient averaging or FedAvg (ablation)FedGMR · CIFAR-10 and ImageNet-100accuracyImproved
add GMR under gradient averaging aggregation (ablation)GA aggregation · CIFAR-10 and ImageNet-100accuracyHurt
use MaskFedAvg without GMR (fixed density) instead of gradient averaging (ablation)MaskFedAvg · ResNet-18 / deeper modelsaccuracyHurt
use naive FedAvg aggregation with zeros for missing coordinates (ablation)FedAvg aggregation · CIFAR-10accuracy: near-chance 10%Hurt
FedGMR vs Fiarse (High heterogeneity IID)FedGMR · ImageNet-100test accuracy: 54.00 -> 60.84Improved
FedGMR vs Fiarse (High heterogeneity Non-IID)FedGMR · ImageNet-100test accuracy: 48.27 -> 58.01Improved
FedGMR vs FedRolex (High heterogeneity IID)FedGMR · ImageNet-100test accuracy: 32.17 -> 60.84Improved
FedGMR vs FedRolex (High heterogeneity Non-IID)FedGMR · ImageNet-100test accuracy: 15.26 -> 58.01Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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