ElaLoRA: Elastic & Learnable Low-Rank Adaptation for Efficient Model Fine-Tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chang, Huandong, Ma, Zicheng, Ma, Mingyuan, Qi, Zhenting, Sabot, Andrew, Jiang, Hong, Kung, H. T.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913769232269312
author Chang, Huandong
Ma, Zicheng
Ma, Mingyuan
Qi, Zhenting
Sabot, Andrew
Jiang, Hong
Kung, H. T.
author_facet Chang, Huandong
Ma, Zicheng
Ma, Mingyuan
Qi, Zhenting
Sabot, Andrew
Jiang, Hong
Kung, H. T.
contents Low-Rank Adaptation (LoRA) has become a widely adopted technique for fine-tuning large-scale pre-trained models with minimal parameter updates. However, existing methods rely on fixed ranks or focus solely on either rank pruning or expansion, failing to adapt ranks dynamically to match the importance of different layers during training. In this work, we propose ElaLoRA, an adaptive low-rank adaptation framework that dynamically prunes and expands ranks based on gradient-derived importance scores. To the best of our knowledge, ElaLoRA is the first method that enables both rank pruning and expansion during fine-tuning. Experiments across multiple benchmarks demonstrate that ElaLoRA consistently outperforms existing PEFT methods across different parameter budgets. Furthermore, our studies validate that layers receiving higher rank allocations contribute more significantly to model performance, providing theoretical justification for our adaptive strategy. By introducing a principled and adaptive rank allocation mechanism, ElaLoRA offers a scalable and efficient fine-tuning solution, particularly suited for resource-constrained environments.
format Preprint
id arxiv_https___arxiv_org_abs_2504_00254
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ElaLoRA: Elastic & Learnable Low-Rank Adaptation for Efficient Model Fine-Tuning
Chang, Huandong
Ma, Zicheng
Ma, Mingyuan
Qi, Zhenting
Sabot, Andrew
Jiang, Hong
Kung, H. T.
Machine Learning
Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
Low-Rank Adaptation (LoRA) has become a widely adopted technique for fine-tuning large-scale pre-trained models with minimal parameter updates. However, existing methods rely on fixed ranks or focus solely on either rank pruning or expansion, failing to adapt ranks dynamically to match the importance of different layers during training. In this work, we propose ElaLoRA, an adaptive low-rank adaptation framework that dynamically prunes and expands ranks based on gradient-derived importance scores. To the best of our knowledge, ElaLoRA is the first method that enables both rank pruning and expansion during fine-tuning. Experiments across multiple benchmarks demonstrate that ElaLoRA consistently outperforms existing PEFT methods across different parameter budgets. Furthermore, our studies validate that layers receiving higher rank allocations contribute more significantly to model performance, providing theoretical justification for our adaptive strategy. By introducing a principled and adaptive rank allocation mechanism, ElaLoRA offers a scalable and efficient fine-tuning solution, particularly suited for resource-constrained environments.
title ElaLoRA: Elastic & Learnable Low-Rank Adaptation for Efficient Model Fine-Tuning
topic Machine Learning
Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.00254