RankMean: Module-Level Importance Score for Merging Fine-tuned LLM Models

Gabriel Perin, Xuxi Chen, Shusen Liu, Bhavya Kailkhura, Zhangyang Wang, Brian Gallagher


Abstract
Traditionally, developing new language models (LMs) capable of addressing multiple tasks involves fine-tuning pre-trained LMs using a wide collection of datasets, a process that often incurs significant computational expenses. Model merging emerges as a cost-effective alternative, allowing the integration of existing models fine-tuned on different tasks into a single model that performs well across all tasks, eliminating the need for additional training. In this paper, we propose RankMean, an algorithm for merging fine-tuned LMs without requiring any downstream data. RankMean determines merging coefficients based on the relative rankings of weight change magnitudes and applies these coefficients for module-wise integration of various fine-tuned models. Our experimental results demonstrate that RankMean outperforms existing baseline methods on multiple benchmarks. The code is available at https://github.com/VITA-Group/RankMean.
Anthology ID:
2024.findings-acl.104
Volume:
Findings of the Association for Computational Linguistics: ACL 2024
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1776–1782
Language:
URL:
https://aclanthology.org/2024.findings-acl.104
DOI:
10.18653/v1/2024.findings-acl.104
Bibkey:
Cite (ACL):
Gabriel Perin, Xuxi Chen, Shusen Liu, Bhavya Kailkhura, Zhangyang Wang, and Brian Gallagher. 2024. RankMean: Module-Level Importance Score for Merging Fine-tuned LLM Models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 1776–1782, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
RankMean: Module-Level Importance Score for Merging Fine-tuned LLM Models (Perin et al., Findings 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.findings-acl.104.pdf