Adapt Once, Thrive with Updates: Transferable Parameter-Efficient Fine-Tuning on Evolving Base Models

TL;DR

This paper introduces Trans-PEFT, a method that improves parameter-efficient fine-tuning by focusing on task-specific patterns, enabling modules to adapt to updated base models without re-tuning, thus reducing maintenance costs.

Contribution

The paper presents Trans-PEFT, a novel approach that enhances PEFT modules to maintain performance after base model updates without re-tuning, supported by theoretical analysis and extensive experiments.

Findings

Trans-PEFT maintains performance on updated models without re-tuning.

Base model updates mainly affect task-specific knowledge in FFN layers.

Trans-PEFT reduces computational costs in model maintenance.

Abstract

Parameter-efficient fine-tuning (PEFT) has become a common method for fine-tuning large language models, where a base model can serve multiple users through PEFT module switching. To enhance user experience, base models require periodic updates. However, once updated, PEFT modules fine-tuned on previous versions often suffer substantial performance degradation on newer versions. Re-tuning these numerous modules to restore performance would incur significant computational costs. Through a comprehensive analysis of the changes that occur during base model updates, we uncover an interesting phenomenon: continual training primarily affects task-specific knowledge stored in Feed-Forward Networks (FFN), while having less impact on the task-specific pattern in the Attention mechanism. Based on these findings, we introduce Trans-PEFT, a novel approach that enhances the PEFT module by focusing on the task-specific pattern while reducing its dependence on certain knowledge in the base model. Further theoretical analysis supports our approach. Extensive experiments across 7 base models and 12 datasets demonstrate that Trans-PEFT trained modules can maintain performance on updated base models without re-tuning, significantly reducing maintenance overhead in real-world applications.