Disperse-Then-Merge: Pushing the Limits of Instruction Tuning via Alignment Tax Reduction

Tingchen Fu, Deng Cai, Lemao Liu, Shuming Shi, Rui Yan


Abstract
Supervised fine-tuning (SFT) on instruction-following corpus is a crucial approach toward the alignment of large language models (LLMs). However, the performance of LLMs on standard knowledge and reasoning benchmarks tends to suffer from deterioration at the latter stage of the SFT process, echoing the phenomenon of alignment tax. Through our pilot study, we put a hypothesis that the data biases are probably one cause behind the phenomenon. To address the issue, we introduce a simple disperse-then-merge framework. To be concrete, we disperse the instruction-following data into portions and then train multiple sub-models using different data portions. Lastly, we merge multiple models into a single one via model merging techniques. Despite its simplicity, our framework outperforms various sophisticated methods such as data curation and training regularization on a series of standard knowledge and reasoning benchmarks.
Anthology ID:
2024.findings-acl.175
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:
2967–2985
Language:
URL:
https://aclanthology.org/2024.findings-acl.175
DOI:
10.18653/v1/2024.findings-acl.175
Bibkey:
Cite (ACL):
Tingchen Fu, Deng Cai, Lemao Liu, Shuming Shi, and Rui Yan. 2024. Disperse-Then-Merge: Pushing the Limits of Instruction Tuning via Alignment Tax Reduction. In Findings of the Association for Computational Linguistics: ACL 2024, pages 2967–2985, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
Disperse-Then-Merge: Pushing the Limits of Instruction Tuning via Alignment Tax Reduction (Fu et al., Findings 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.findings-acl.175.pdf