Activated Parameter Locating via Causal Intervention for Model Merging

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kong, Fanshuang, Zhang, Richong, Wang, Ziqiao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914915576446976
author Kong, Fanshuang
Zhang, Richong
Wang, Ziqiao
author_facet Kong, Fanshuang
Zhang, Richong
Wang, Ziqiao
contents Model merging combines multiple homologous models into one model, achieving convincing generalization without the necessity of additional training. A key challenge in this problem is resolving parameter redundancies and conflicts across multiple models. Existing models have demonstrated that dropping a portion of delta parameters can alleviate conflicts while maintaining performance. However, these methods often drop parameters either randomly or based on magnitude, overlooking task-specific information embedded in fine-tuned models. In this paper, we propose an Activated Parameter Locating (APL) method that utilizes causal intervention to estimate parameter importance, enabling more precise parameter drops and better conflict mitigation. Moreover, to reduce the computational complexity associated with a large number of parameter partitions, we also introduce a theoretically supported gradient approximation strategy for APL. Experiments on model merging within both in-domain and out-of-domain settings, along with associated analyses, showcase the effectiveness of APL.
format Preprint
id arxiv_https___arxiv_org_abs_2408_09485
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Activated Parameter Locating via Causal Intervention for Model Merging
Kong, Fanshuang
Zhang, Richong
Wang, Ziqiao
Computation and Language
Model merging combines multiple homologous models into one model, achieving convincing generalization without the necessity of additional training. A key challenge in this problem is resolving parameter redundancies and conflicts across multiple models. Existing models have demonstrated that dropping a portion of delta parameters can alleviate conflicts while maintaining performance. However, these methods often drop parameters either randomly or based on magnitude, overlooking task-specific information embedded in fine-tuned models. In this paper, we propose an Activated Parameter Locating (APL) method that utilizes causal intervention to estimate parameter importance, enabling more precise parameter drops and better conflict mitigation. Moreover, to reduce the computational complexity associated with a large number of parameter partitions, we also introduce a theoretically supported gradient approximation strategy for APL. Experiments on model merging within both in-domain and out-of-domain settings, along with associated analyses, showcase the effectiveness of APL.
title Activated Parameter Locating via Causal Intervention for Model Merging
topic Computation and Language
url https://arxiv.org/abs/2408.09485