Enhancing Large Language Model Performance with Gradient-Based Parameter Selection

TL;DR

This paper introduces Gradient-Mask Tuning (GMT), a novel method that selectively updates parameters based on gradient magnitudes, improving large language model fine-tuning efficiency and performance across various tasks.

Contribution

GMT leverages gradient information to identify important parameters for task-specific fine-tuning, outperforming traditional methods and increasing LLM capabilities.

Findings

GMT outperforms traditional fine-tuning methods.

GMT is insensitive to mask ratio.

GMT has computational efficiency comparable to vanilla SFT.

Abstract

Large language models (LLMs) have revolutionized lots of fields of research. Although it is well-known that fine-tuning is essential for enhancing the capabilities of LLMs, existing research suggests that there is potential redundancy in the fine-tuning process and therefore proposes to update only a subset of parameters. However, these methods fail to leverage the task-specific information to identify important parameters during training. Based on the insight that gradients inherently contain information on task-specific data, we propose Gradient-Mask Tuning (GMT), a method that selectively updates parameters during training based on their gradient information. Specifically, we compute the absolute values of the gradients and apply masking to those with relatively smaller magnitudes. Our empirical results across various tasks demonstrate that GMT not only outperforms traditional fine-tuning methods but also elevates the upper limits of LLM performance. Further analysis indicates that GMT exhibits insensitivity to mask ratio and possesses computational efficiency comparable to vanilla SFT.