Finetune Once: Decoupling General & Domain Learning with Dynamic Boosted Annealing

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tang, Yang, Liu, Ruijie, Wang, Yifan, Li, Shiyu, Chen, Xi
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912653464567808
author Tang, Yang
Liu, Ruijie
Wang, Yifan
Li, Shiyu
Chen, Xi
author_facet Tang, Yang
Liu, Ruijie
Wang, Yifan
Li, Shiyu
Chen, Xi
contents Large language models (LLMs) fine-tuning shows excellent implications. However, vanilla fine-tuning methods often require intricate data mixture and repeated experiments for optimal generalization. To address these challenges and streamline the training process, we propose an efficient and universal solution, Dynamic Boosted Annealing (DBA). We obtain a global gradient through zero-learning-rate training on general data, which is subsequently employed for gradient boosting and dynamic training step correction during domain training. In conjunction with annealing learning, we end up establishing a fine-tuning pipeline that relies solely on domain data without collapse. By evaluating both general and domain-specific performance across multiple tasks on several popular base models, DBA achieves an average improvement of 5.8% in joint performance over vanilla fine-tuning. Furthermore, since general data is no longer involved in annealing, repeated experiments led by data mixture are also eliminated. According to our tests, the DBA method can reduce GPU hours by 91.0% compared to the vanilla method.
format Preprint
id arxiv_https___arxiv_org_abs_2509_26242
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Finetune Once: Decoupling General & Domain Learning with Dynamic Boosted Annealing
Tang, Yang
Liu, Ruijie
Wang, Yifan
Li, Shiyu
Chen, Xi
Computation and Language
Artificial Intelligence
Machine Learning
Large language models (LLMs) fine-tuning shows excellent implications. However, vanilla fine-tuning methods often require intricate data mixture and repeated experiments for optimal generalization. To address these challenges and streamline the training process, we propose an efficient and universal solution, Dynamic Boosted Annealing (DBA). We obtain a global gradient through zero-learning-rate training on general data, which is subsequently employed for gradient boosting and dynamic training step correction during domain training. In conjunction with annealing learning, we end up establishing a fine-tuning pipeline that relies solely on domain data without collapse. By evaluating both general and domain-specific performance across multiple tasks on several popular base models, DBA achieves an average improvement of 5.8% in joint performance over vanilla fine-tuning. Furthermore, since general data is no longer involved in annealing, repeated experiments led by data mixture are also eliminated. According to our tests, the DBA method can reduce GPU hours by 91.0% compared to the vanilla method.
title Finetune Once: Decoupling General & Domain Learning with Dynamic Boosted Annealing
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2509.26242