AI Paper Atlas

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

BESplit: Bias-Compensated Split Federated Learning with Evidential Aggregation

Yuhan Xie, Chen Lyu, Jingrong Huang

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning · 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
add Bias-Compensated Collaboration (BCC) (ablation)BESplit-G · CIFAR10 (non-IID)accuracy: 55.65 -> 59.82Improved
add BCC (ablation)BESplit-G · ISIC (non-IID)accuracy: 63.59 -> 67.45Improved
add BCC (ablation)BESplit-G · CIFAR100 (non-IID)accuracy: 64.28 -> 68.17Improved
add Evidential Aggregation (EA) instead of simple average (ablation)BESplit-G · ISIC (non-IID)accuracy: 64.34 -> 67.45Improved
add EA (ablation)BESplit-G · CIFAR10 (non-IID)accuracy: 57.49 -> 59.82Improved
add evidence concentration term E in EA (ablation)BESplit-G · CIFAR100 (non-IID)accuracy: 66.42 -> 68.17Improved
add aleatoric uncertainty penalty U_ale in EA (ablation)BESplit-G · CIFAR100 (non-IID)accuracy: 67.23 -> 68.17Improved
add epistemic uncertainty penalty U_epi in EA (ablation)BESplit-G · CIFAR100 (non-IID)accuracy: 67.75 -> 68.17Improved
add U_epi in EA (ablation)BESplit-G · F-MNIST (non-IID)accuracy: 74.11 -> 74.89Improved
use BESplit-G instead of best FL baseline FAVDBESplit-G · HAM10000 (non-IID)accuracy: 67.90 -> 70.46Improved
use full BESplit instead of FedAvg for communication roundsBESplit · CIFAR100 non-IID 60% targetrounds: 475 -> 371Improved
use BESplit-G instead of FedAvg for training timeBESplit-G · HAM10000speedup: 1.58xImproved

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