Mitigating Training Imbalance in LLM Fine-Tuning via Selective Parameter Merging

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ju, Yiming, Ni, Ziyi, Xing, Xingrun, Zeng, Zhixiong, Zhao, hanyu, Fan, Siqi, Zhang, Zheng
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917795347824640
author Ju, Yiming
Ni, Ziyi
Xing, Xingrun
Zeng, Zhixiong
Zhao, hanyu
Fan, Siqi
Zhang, Zheng
author_facet Ju, Yiming
Ni, Ziyi
Xing, Xingrun
Zeng, Zhixiong
Zhao, hanyu
Fan, Siqi
Zhang, Zheng
contents Supervised fine-tuning (SFT) is crucial for adapting Large Language Models (LLMs) to specific tasks. In this work, we demonstrate that the order of training data can lead to significant training imbalances, potentially resulting in performance degradation. Consequently, we propose to mitigate this imbalance by merging SFT models fine-tuned with different data orders, thereby enhancing the overall effectiveness of SFT. Additionally, we introduce a novel technique, "parameter-selection merging," which outperforms traditional weighted-average methods on five datasets. Further, through analysis and ablation studies, we validate the effectiveness of our method and identify the sources of performance improvements.
format Preprint
id arxiv_https___arxiv_org_abs_2410_03743
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Mitigating Training Imbalance in LLM Fine-Tuning via Selective Parameter Merging
Ju, Yiming
Ni, Ziyi
Xing, Xingrun
Zeng, Zhixiong
Zhao, hanyu
Fan, Siqi
Zhang, Zheng
Computation and Language
Artificial Intelligence
Machine Learning
Supervised fine-tuning (SFT) is crucial for adapting Large Language Models (LLMs) to specific tasks. In this work, we demonstrate that the order of training data can lead to significant training imbalances, potentially resulting in performance degradation. Consequently, we propose to mitigate this imbalance by merging SFT models fine-tuned with different data orders, thereby enhancing the overall effectiveness of SFT. Additionally, we introduce a novel technique, "parameter-selection merging," which outperforms traditional weighted-average methods on five datasets. Further, through analysis and ablation studies, we validate the effectiveness of our method and identify the sources of performance improvements.
title Mitigating Training Imbalance in LLM Fine-Tuning via Selective Parameter Merging
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2410.03743