AI Paper Atlas

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

Beyond Aggregation: Guiding Clients in Heterogeneous Federated Learning

Zijian Wang, Xiaofei Zhang, Xin Zhang, Yukun Liu, Qiong Zhang

ICLR 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated learning with pretrained models

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 client-guidance head with density ratio / empirical likelihood losspersonalized FL with ResNet-18 · CIFAR-10 Dir-0.3System accuracy: 54.90 -> 62.78 (vs FedAvgFT)Improved
add client-guidance headpersonalized FL with ResNet-18 · CIFAR-100 25-SPCSystem accuracy: 27.96 -> 31.24 (vs FedAvgFT)Improved
add client-guidance headpersonalized FL with ResNet-18 · CIFAR-20 Dir-0.3System accuracy: 37.53 -> 37.63 (vs FedAvgFT)No clear effect
add client-guidance headpersonalized FL with ResNet-18 · CIFAR-100 25-SPCAverage accuracy: 44.26 -> 46.73 (vs FedAS)Improved
use label shift 5-SPC instead of Dir-0.3all methods · CIFAR-10accuracy: 62.78 -> 58.50 (ours, system)Hurt
increase lambda weight toward image classification (ablation)guidance framework · CIFAR-10 Dir-0.3System accuracy vs client accuracyMixed
raise covariate shift intensity (ablation)guidance framework · CIFAR-10 Dir-0.3accuracyMixed
change backbone sharing strategy (none, shallow, mid, deep) (ablation)guidance framework · CIFAR-10 Dir-0.3 (LeNet)accuracyNo clear effect
increase number of clients from 8 to 32 (ablation)guidance framework · CIFAR-10System accuracy: 59.59 -> 46.62Hurt
add client guidance on real medical dataguidance framework · RETINASystem accuracy: +1.41 to +7.67 points over baselinesImproved

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