FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Changyu, Huang, Shuanghong, Liu, Jiashen, Lei, Ming, Xing, Jidu, Wu, Kaishun, Wang, Lu, Luo, Fei
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910173308649472
author Li, Changyu
Huang, Shuanghong
Liu, Jiashen
Lei, Ming
Xing, Jidu
Wu, Kaishun
Wang, Lu
Luo, Fei
author_facet Li, Changyu
Huang, Shuanghong
Liu, Jiashen
Lei, Ming
Xing, Jidu
Wu, Kaishun
Wang, Lu
Luo, Fei
contents Federated fine-tuning provides a practical route to adapt large language models (LLMs) on edge devices without centralizing private data, yet in mobile deployments the training wall-clock is often bottlenecked by straggler-limited uplink communication under heterogeneous bandwidth and intermittent participation. Although parameter-efficient fine-tuning (PEFT) reduces trainable parameters, per-round payloads remain prohibitive in non-IID regimes, where uniform compression can discard rare but task-critical signals. We propose Fed-FSTQ, a Fisher-guided token quantization system primitive for communication-efficient federated LLM fine-tuning. Fed-FSTQ employs a lightweight Fisher proxy to estimate token sensitivity, coupling importance-aware token selection with non-uniform mixed-precision quantization to allocate higher fidelity to informative evidence while suppressing redundant transmission. The method is model-agnostic, serves as a drop-in module for standard federated PEFT pipelines, e.g., LoRA, without modifying the server aggregation rule, and supports bandwidth-heterogeneous clients via compact sparse message packing. Experiments on multilingual QA and medical QA under non-IID partitions show that Fed-FSTQ reduces cumulative uplink traffic required to reach a fixed quality threshold by 46x relative to a standard LoRA baseline, and improves end-to-end wall-clock time-to-accuracy by 52%. Furthermore, enabling Fisher-guided token reduction at inference yields up to a 1.55x end-to-end speedup on NVIDIA Jetson-class edge devices, demonstrating deployability under tight resource constraints.
format Preprint
id arxiv_https___arxiv_org_abs_2604_25421
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices
Li, Changyu
Huang, Shuanghong
Liu, Jiashen
Lei, Ming
Xing, Jidu
Wu, Kaishun
Wang, Lu
Luo, Fei
Machine Learning
Artificial Intelligence
C.2.4; I.2.7
Federated fine-tuning provides a practical route to adapt large language models (LLMs) on edge devices without centralizing private data, yet in mobile deployments the training wall-clock is often bottlenecked by straggler-limited uplink communication under heterogeneous bandwidth and intermittent participation. Although parameter-efficient fine-tuning (PEFT) reduces trainable parameters, per-round payloads remain prohibitive in non-IID regimes, where uniform compression can discard rare but task-critical signals. We propose Fed-FSTQ, a Fisher-guided token quantization system primitive for communication-efficient federated LLM fine-tuning. Fed-FSTQ employs a lightweight Fisher proxy to estimate token sensitivity, coupling importance-aware token selection with non-uniform mixed-precision quantization to allocate higher fidelity to informative evidence while suppressing redundant transmission. The method is model-agnostic, serves as a drop-in module for standard federated PEFT pipelines, e.g., LoRA, without modifying the server aggregation rule, and supports bandwidth-heterogeneous clients via compact sparse message packing. Experiments on multilingual QA and medical QA under non-IID partitions show that Fed-FSTQ reduces cumulative uplink traffic required to reach a fixed quality threshold by 46x relative to a standard LoRA baseline, and improves end-to-end wall-clock time-to-accuracy by 52%. Furthermore, enabling Fisher-guided token reduction at inference yields up to a 1.55x end-to-end speedup on NVIDIA Jetson-class edge devices, demonstrating deployability under tight resource constraints.
title FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices
topic Machine Learning
Artificial Intelligence
C.2.4; I.2.7
url https://arxiv.org/abs/2604.25421