AI Paper Atlas

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

Prototype-guided Bilateral Alignment Multimodal Federated Learning

Tianchi Liao, Lele Fu, Sheng Huang, Qing Hu, Hong-Ning Dai, Chuan Chen

ICML 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 MFedPBA bilateral prototype alignment vs best baselineheterogeneous multimodal FL · dataset block 1 (M2)test accuracy (%): 45.29 -> 49.04Improved
naive prototype aggregation (FedProto) instead of local trainingheterogeneous multimodal FL · dataset block 1 (M2)test accuracy (%): 42.65 -> 38.72Hurt
remove reconstruction loss L_rec on server (ablation)MFedPBA · Caltech101 (M2)test accuracy (%)Hurt
remove any server-side module (contrastive, GW alignment) (ablation)MFedPBA · Caltech101 (M2)test accuracy (%)Hurt
replace entropy-weighted logit aggregation with equal averaging (ablation)MFedPBA · Caltech101 (M2)test accuracy (%)Hurt
increase feature dimension from 48 to 256 (ablation)MFedPBA · Youtube (M1+)test accuracy (%): 47.02 -> 43.84Hurt
reduce feature dimension from 48 to 24 (ablation)MFedPBA · Youtube (M1+)test accuracy (%): 47.02 -> 44.43Hurt
increase server epochs S from 50 to 1000 (ablation)MFedPBA · server epochs sensitivitytest accuracy (%)No clear effect
use very large loss weights lambda1, lambda2 (ablation)MFedPBA · Youtubetest accuracy (%)Hurt
scale to more clients (K=50, 100, 20)MFedPBA · large-client settingstest accuracy (%): 30.88 -> 31.19Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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