@Article{cmc.2026.086477,
AUTHOR = {Hongzhi Li, Jiale Wu, Dun Li, Kezhong Lu, Qishou Xia},
TITLE = {An Intelligent Lesson Preparation Assistant System Using a Fine-Tuned Large Language Model for Computer Science Education},
JOURNAL = {Computers, Materials \& Continua},
VOLUME = {},
YEAR = {},
NUMBER = {},
PAGES = {{pages}},
URL = {http://www.techscience.com/cmc/online/detail/28350},
ISSN = {1546-2226},
ABSTRACT = {The application of large language models (LLMs) to computer science (CS) education has great potential for intelligent lesson preparation, yet existing systems remain limited by weak textbook alignment, inconsistent pedagogical structures, and extensive manual post-editing. This paper presents the Fine-Tuned DeepSeek-driven Intelligent Lesson Preparation Assistant System (FT-DILPAS), a textbook-driven lesson preparation framework that integrates efficient model adaptation, structured textbook understanding, curriculum-aware prompting, and retrieval-augmented generation into a unified pipeline. Evaluations on 60 CS textbooks show that FT-DILPAS achieves parsing accuracies of 96.5% for standard layouts and 88.2% for complex layouts, significantly outperforming traditional parsing methods. Human evaluations further demonstrate improvements of 50.8% in format standardization and 30.6% in textbook alignment over GPT-4o, with consistently superior pedagogical quality across all evaluated general-purpose LLMs. Overall, FT-DILPAS provides a practical and scalable paradigm for curriculum-aligned, AI-assisted lesson preparation in computer science education.},
DOI = {10.32604/cmc.2026.086477}
}