EDCO: Dynamic Curriculum Orchestration for Domain-specific Large Language Model Fine-tuning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Pang, Jing-Cheng, Sun, Liu, Zhou, Chang, Tang, Xian, Ma, Haichuan, Jiang, Kun, Wang, Jianlong, Zhang, Kai, Wu, Sijie, Cai, Haoran, Wu, Chenwei, Li, Xubin, Chen, Xin
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914238191108096
author Pang, Jing-Cheng
Sun, Liu
Zhou, Chang
Tang, Xian
Ma, Haichuan
Jiang, Kun
Wang, Jianlong
Zhang, Kai
Wu, Sijie
Cai, Haoran
Wu, Chenwei
Li, Xubin
Chen, Xin
author_facet Pang, Jing-Cheng
Sun, Liu
Zhou, Chang
Tang, Xian
Ma, Haichuan
Jiang, Kun
Wang, Jianlong
Zhang, Kai
Wu, Sijie
Cai, Haoran
Wu, Chenwei
Li, Xubin
Chen, Xin
contents Domain-specific large language models (LLMs), typically developed by fine-tuning a pre-trained general-purpose LLM on specialized datasets, represent a significant advancement in applied AI. A common strategy in LLM fine-tuning is curriculum learning, which pre-orders training samples based on metrics like difficulty to improve learning efficiency compared to a random sampling strategy. However, most existing methods for LLM fine-tuning rely on a static curriculum, designed prior to training, which lacks adaptability to the model's evolving needs during fine-tuning. To address this, we propose EDCO, a novel framework based on two key concepts: inference entropy and dynamic curriculum orchestration. Inspired by recent findings that maintaining high answer entropy benefits long-term reasoning gains, EDCO prioritizes samples with high inference entropy in a continuously adapted curriculum. EDCO integrates three core components: an efficient entropy estimator that uses prefix tokens to approximate full-sequence entropy, an entropy-based curriculum generator that selects data points with the highest inference entropy, and an LLM trainer that optimizes the model on the selected curriculum. Comprehensive experiments in communication, medicine and law domains, EDCO outperforms traditional curriculum strategies for fine-tuning Qwen3-4B and Llama3.2-3B models under supervised and reinforcement learning settings. Furthermore, the proposed efficient entropy estimation reduces computational time by 83.5% while maintaining high accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2601_03725
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle EDCO: Dynamic Curriculum Orchestration for Domain-specific Large Language Model Fine-tuning
Pang, Jing-Cheng
Sun, Liu
Zhou, Chang
Tang, Xian
Ma, Haichuan
Jiang, Kun
Wang, Jianlong
Zhang, Kai
Wu, Sijie
Cai, Haoran
Wu, Chenwei
Li, Xubin
Chen, Xin
Machine Learning
Domain-specific large language models (LLMs), typically developed by fine-tuning a pre-trained general-purpose LLM on specialized datasets, represent a significant advancement in applied AI. A common strategy in LLM fine-tuning is curriculum learning, which pre-orders training samples based on metrics like difficulty to improve learning efficiency compared to a random sampling strategy. However, most existing methods for LLM fine-tuning rely on a static curriculum, designed prior to training, which lacks adaptability to the model's evolving needs during fine-tuning. To address this, we propose EDCO, a novel framework based on two key concepts: inference entropy and dynamic curriculum orchestration. Inspired by recent findings that maintaining high answer entropy benefits long-term reasoning gains, EDCO prioritizes samples with high inference entropy in a continuously adapted curriculum. EDCO integrates three core components: an efficient entropy estimator that uses prefix tokens to approximate full-sequence entropy, an entropy-based curriculum generator that selects data points with the highest inference entropy, and an LLM trainer that optimizes the model on the selected curriculum. Comprehensive experiments in communication, medicine and law domains, EDCO outperforms traditional curriculum strategies for fine-tuning Qwen3-4B and Llama3.2-3B models under supervised and reinforcement learning settings. Furthermore, the proposed efficient entropy estimation reduces computational time by 83.5% while maintaining high accuracy.
title EDCO: Dynamic Curriculum Orchestration for Domain-specific Large Language Model Fine-tuning
topic Machine Learning
url https://arxiv.org/abs/2601.03725