Revisiting LoRA through the Lens of Parameter Redundancy: Spectral Encoding Helps

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cheng, Jiashun, Chen, Aochuan, Chen, Nuo, Gao, Ziqi, Li, Yuhan, Li, Jia, Tsung, Fugee
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912442503659520
author Cheng, Jiashun
Chen, Aochuan
Chen, Nuo
Gao, Ziqi
Li, Yuhan
Li, Jia
Tsung, Fugee
author_facet Cheng, Jiashun
Chen, Aochuan
Chen, Nuo
Gao, Ziqi
Li, Yuhan
Li, Jia
Tsung, Fugee
contents Low-Rank Adaptation (LoRA) has emerged as a prominent technique for fine-tuning large foundation models. Despite its successes, the substantial parameter redundancy, which limits the capacity and efficiency of LoRA, has been recognized as a bottleneck. In this work, we systematically investigate the impact of redundancy in fine-tuning LoRA and reveal that reducing density redundancy does not degrade expressiveness. Based on this insight, we introduce \underline{S}pectral-\underline{e}ncoding \underline{L}ow-\underline{R}ank \underline{A}daptation (SeLoRA), which harnesses the robust expressiveness of spectral bases to re-parameterize LoRA from a sparse spectral subspace. Designed with simplicity, SeLoRA enables seamless integration with various LoRA variants for performance boosting, serving as a scalable plug-and-play framework. Extensive experiments substantiate that SeLoRA achieves greater efficiency with fewer parameters, delivering superior performance enhancements over strong baselines on various downstream tasks, including commonsense reasoning, math reasoning, and code generation.
format Preprint
id arxiv_https___arxiv_org_abs_2506_16787
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Revisiting LoRA through the Lens of Parameter Redundancy: Spectral Encoding Helps
Cheng, Jiashun
Chen, Aochuan
Chen, Nuo
Gao, Ziqi
Li, Yuhan
Li, Jia
Tsung, Fugee
Machine Learning
Low-Rank Adaptation (LoRA) has emerged as a prominent technique for fine-tuning large foundation models. Despite its successes, the substantial parameter redundancy, which limits the capacity and efficiency of LoRA, has been recognized as a bottleneck. In this work, we systematically investigate the impact of redundancy in fine-tuning LoRA and reveal that reducing density redundancy does not degrade expressiveness. Based on this insight, we introduce \underline{S}pectral-\underline{e}ncoding \underline{L}ow-\underline{R}ank \underline{A}daptation (SeLoRA), which harnesses the robust expressiveness of spectral bases to re-parameterize LoRA from a sparse spectral subspace. Designed with simplicity, SeLoRA enables seamless integration with various LoRA variants for performance boosting, serving as a scalable plug-and-play framework. Extensive experiments substantiate that SeLoRA achieves greater efficiency with fewer parameters, delivering superior performance enhancements over strong baselines on various downstream tasks, including commonsense reasoning, math reasoning, and code generation.
title Revisiting LoRA through the Lens of Parameter Redundancy: Spectral Encoding Helps
topic Machine Learning
url https://arxiv.org/abs/2506.16787