AI Paper Atlas

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

UniFLoW: Universal Multi-Modal Federated LoRA Fine-Tuning Framework with Analytical Aggregation

Haoyuan Liang, Zhiyu Ye, Jielong Tang, Yang Yang, Shilei Cao, Guowen Li, Fei Hu, Zhiwei Zhang, Haohuan Fu, Juepeng Zheng

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated prompt and LoRA tuning

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
replace FedSA-LoRA aggregation with FedA2-LoRALoRA on RoBERTa · GLUE avgaccuracy: 90.39 -> 90.50Improved
replace rsLoRA aggregation with FedA2-rsLoRArsLoRA on RoBERTa · GLUE avgaccuracy: 89.16 -> 90.68Improved
replace VeRA aggregation with FedA2-VeRAVeRA on RoBERTa · GLUE avgaccuracy: 86.86 -> 88.02Improved
replace LoRA aggregation with FedA2-LoRALoRA · PandaGPT image test, image+audio clientsAcc: 64.89 -> 75.00Improved
add II stage training (encoder first, then LLM) (ablation)LoRA · PandaGPT image test, image-only clientsAcc: 64.68 -> 66.67Improved
add II stage training (ablation)FFA-LoRA · PandaGPT image test, image+audio clientsAcc: 63.15 -> 70.58Improved
add II stage training (ablation)FedA2-LoRA · PandaGPT image test, image-only clientsAcc: 69.02 -> 72.22Improved
add II stage training (ablation)FedSA-LoRA · HeySQuAD audio test, audio-only clientsAcc: 53.84 -> 53.84No clear effect
add Tikhonov regularization (lambda=0.1) to B recovery (ablation)FedA2-LoRA · PandaGPT image test, image+audio clientsAcc: 65.09 -> 76.19Improved
add Tikhonov regularization to B recovery (ablation)FedA2-LoRA · PandaGPT image test, image+audio clientsFBERT: 80.96 -> 85.69Improved
use FedA2-LoRA with heterogeneous ranks (zero-padding) instead of LoRALoRA · PandaGPT image test, heterogeneous-rank clientsFBERT: 82.34 -> 84.46Improved
replace LoRA aggregation with FedA2-LoRA using LLaMA-1B and UniBindLoRA · PandaGPT image test, LLaMA-1B + UniBindAcc: 63.15 -> 65.00Improved

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