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

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Gu, Naibin, Fu, Peng, Liu, Xiyu, Ma, Ke, Lin, Zheng, Wang, Weiping
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910994914082816
author Gu, Naibin
Fu, Peng
Liu, Xiyu
Ma, Ke
Lin, Zheng
Wang, Weiping
author_facet Gu, Naibin
Fu, Peng
Liu, Xiyu
Ma, Ke
Lin, Zheng
Wang, Weiping
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2506_06844
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Adapt Once, Thrive with Updates: Transferable Parameter-Efficient Fine-Tuning on Evolving Base Models
Gu, Naibin
Fu, Peng
Liu, Xiyu
Ma, Ke
Lin, Zheng
Wang, Weiping
Computation and Language
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.
title Adapt Once, Thrive with Updates: Transferable Parameter-Efficient Fine-Tuning on Evolving Base Models
topic Computation and Language
url https://arxiv.org/abs/2506.06844