English

Localized LoRA: A Structured Low-Rank Approximation for Efficient Fine-Tuning

Machine Learning 2025-09-25 v2 Artificial Intelligence Computation and Language

Abstract

Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, offer compact and effective alternatives to full model fine-tuning by introducing low-rank updates to pre-trained weights. However, most existing approaches rely on global low rank structures, which can overlook spatial patterns spread across the parameter space. In this work, we propose Localized LoRA, a generalized framework that models weight updates as a composition of low-rank matrices applied to structured blocks of the weight matrix. This formulation enables dense, localized updates throughout the parameter space without increasing the total number of trainable parameters. We provide a formal comparison between global, diagonal-local, and fully localized low-rank approximations, and show that our method consistently achieves lower approximation error under matched parameter budgets. Experiments on both synthetic and practical settings demonstrate that Localized LoRA offers a more expressive and adaptable alternative to existing methods, enabling efficient fine-tuning with improved performance.

Keywords

Cite

@article{arxiv.2506.00236,
  title  = {Localized LoRA: A Structured Low-Rank Approximation for Efficient Fine-Tuning},
  author = {Babak Barazandeh and Subhabrata Majumdar and Om Rajyaguru and George Michailidis},
  journal= {arXiv preprint arXiv:2506.00236},
  year   = {2025}
}
R2 v1 2026-07-01T02:51:45.039Z