Saved in:
Bibliographic Details
Main Authors: Wang, Xiaoyu, Li, Xiaotian, Zhou, Zhixiang, Li, Chen, Liu, Yong
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.00451
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918316555108352
author Wang, Xiaoyu
Li, Xiaotian
Zhou, Zhixiang
Li, Chen
Liu, Yong
author_facet Wang, Xiaoyu
Li, Xiaotian
Zhou, Zhixiang
Li, Chen
Liu, Yong
contents Decentralized federated learning (DFL), a serverless variant of federated learning, poses unique challenges for parameter-efficient fine-tuning due to the factorized structure of low-rank adaptation (LoRA). Unlike linear parameters, decentralized aggregation of LoRA updates introduces topology-dependent cross terms that can destabilize training under dynamic communication graphs. We propose \texttt{TAD-LoRA}, a Topology-Aware Decentralized Low-Rank Adaptation framework that coordinates the updates and mixing of LoRA factors to control inter-client misalignment. We theoretically prove the convergence of \texttt{TAD-LoRA} under non-convex objectives, explicitly characterizing the trade-off between topology-induced cross-term error and block-coordinate representation bias governed by the switching interval of alternative training. Experiments under various communication conditions validate our analysis, showing that \texttt{TAD-LoRA} achieves robust performance across different communication scenarios, remaining competitive in strongly connected topologies and delivering clear gains under moderately and weakly connected topologies, with particularly strong results on the MNLI dataset.
format Preprint
id arxiv_https___arxiv_org_abs_2602_00451
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Stabilizing Decentralized Federated Fine-Tuning via Topology-Aware Alternating LoRA
Wang, Xiaoyu
Li, Xiaotian
Zhou, Zhixiang
Li, Chen
Liu, Yong
Machine Learning
Distributed, Parallel, and Cluster Computing
Decentralized federated learning (DFL), a serverless variant of federated learning, poses unique challenges for parameter-efficient fine-tuning due to the factorized structure of low-rank adaptation (LoRA). Unlike linear parameters, decentralized aggregation of LoRA updates introduces topology-dependent cross terms that can destabilize training under dynamic communication graphs. We propose \texttt{TAD-LoRA}, a Topology-Aware Decentralized Low-Rank Adaptation framework that coordinates the updates and mixing of LoRA factors to control inter-client misalignment. We theoretically prove the convergence of \texttt{TAD-LoRA} under non-convex objectives, explicitly characterizing the trade-off between topology-induced cross-term error and block-coordinate representation bias governed by the switching interval of alternative training. Experiments under various communication conditions validate our analysis, showing that \texttt{TAD-LoRA} achieves robust performance across different communication scenarios, remaining competitive in strongly connected topologies and delivering clear gains under moderately and weakly connected topologies, with particularly strong results on the MNLI dataset.
title Stabilizing Decentralized Federated Fine-Tuning via Topology-Aware Alternating LoRA
topic Machine Learning
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2602.00451