Saved in:
Bibliographic Details
Main Authors: Lou, Yuchen, Ye, Zeqi, Chen, Minshuo
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.02216
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911439051030528
author Lou, Yuchen
Ye, Zeqi
Chen, Minshuo
author_facet Lou, Yuchen
Ye, Zeqi
Chen, Minshuo
contents This paper develops a new perspective on parameter-efficient fine-tuning (PEFT) for LLMs, inspired by classical subspace minimization. We introduce a unifying framework, Parameter-Efficient Subspace Optimization (PESO), which recovers existing methods such as LoRA and connects them to the principled algorithmic and theoretical foundations of subspace optimization. This connection highlights a natural ``exploration--exploitation'' view of subspace methods, guiding the design of new algorithms that achieve strong convergence performance while still preserving memory efficiency. We instantiate the framework into a practical algorithm, PESO-LoRA, based on a LoRA-type parameterization. Importantly, we provide convergence guarantees stated in the full-parameter space for the induced update, addressing a key limitation of LoRA-style analyses that only track low-dimensional factors. Empirically, PESO-LoRA improves over strong PEFT baselines on standard fine-tuning benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2512_02216
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Parameter-Efficient Subspace Optimization for LLM Fine-Tuning
Lou, Yuchen
Ye, Zeqi
Chen, Minshuo
Optimization and Control
This paper develops a new perspective on parameter-efficient fine-tuning (PEFT) for LLMs, inspired by classical subspace minimization. We introduce a unifying framework, Parameter-Efficient Subspace Optimization (PESO), which recovers existing methods such as LoRA and connects them to the principled algorithmic and theoretical foundations of subspace optimization. This connection highlights a natural ``exploration--exploitation'' view of subspace methods, guiding the design of new algorithms that achieve strong convergence performance while still preserving memory efficiency. We instantiate the framework into a practical algorithm, PESO-LoRA, based on a LoRA-type parameterization. Importantly, we provide convergence guarantees stated in the full-parameter space for the induced update, addressing a key limitation of LoRA-style analyses that only track low-dimensional factors. Empirically, PESO-LoRA improves over strong PEFT baselines on standard fine-tuning benchmarks.
title Parameter-Efficient Subspace Optimization for LLM Fine-Tuning
topic Optimization and Control
url https://arxiv.org/abs/2512.02216