UORA: Uniform Orthogonal Reinitialization Adaptation in Parameter-Efficient Fine-Tuning of Large Models

TL;DR

UORA is a novel parameter-efficient fine-tuning method for large models that uses a low-rank reinitialization approach to outperform existing techniques in performance and efficiency.

Contribution

It introduces a new reparametrization mechanism that reduces trainable parameters and improves fine-tuning efficiency over prior methods like LoRA and VeRA.

Findings

UORA achieves state-of-the-art performance on benchmarks.

It significantly reduces the number of trainable parameters.

UORA outperforms existing methods in computation and storage efficiency.

Abstract

This paper introduces Uniform Orthogonal Reinitialization Adaptation (UORA), a novel parameter-efficient fine-tuning (PEFT) approach for Large Language Models (LLMs). UORA achieves state-of-the-art performance and parameter efficiency by leveraging a low-rank approximation method to reduce the number of trainable parameters. Unlike existing methods such as LoRA and VeRA, UORA employs an interpolation-based reparametrization mechanism that selectively reinitializes rows and columns in frozen projection matrices, guided by the vector magnitude heuristic. This results in substantially fewer trainable parameters compared to LoRA and outperforms VeRA in computation and storage efficiency. Comprehensive experiments across various benchmarks demonstrate UORA's superiority in achieving competitive fine-tuning performance with negligible computational overhead. We demonstrate its performance on GLUE and E2E benchmarks and its effectiveness in instruction-tuning large language models and image classification models. Our contributions establish a new paradigm for scalable and resource-efficient fine-tuning of LLMs.