情报杂志2025，Vol.44Issue(6)：119-126,118,9.DOI:10.3969/j.issn.1002-1965.2025.06.016
基于大语言模型知识自蒸馏的引用意图自动识别
Automatic Citaion Intention Recognition Based on Knowledge Self-Distillation of Large Language Model
摘要
Abstract
[Research purpose]This paper constructs an automatic citation intention recognition method based on knowledge self-distil-lation technology of large language model to improve the accuracy and generalization of recognition,and provide a basis and reference for realizing more efficient scientific data organization and management.[Research method]This paper,a training method integrating vari-ous fine-tuning techniques is proposed.Specifically,the fine-tuning prompt word template of multiple role strategy is used to improve the generalization of large language model,and the comprehensive performance of the large language model is improved by combining knowl-edge self-distillation technology.[Research result/conclusion]The research shows that the recall rate of citation intention recognition of the large model trained by the fine-tuning method reaches 75.3％,the precision rate reaches 86.0％,and the value reaches the level of 80.3％.All the indicators are better than the baseline model and the existing citation intention recognition model,which can effectively improve the automatic citation intention recognition effect on the basis of reducing the cost.关键词
引用意图/大语言模型/知识自蒸馏/创新性评价/文本分类/科技评价Key words
intention of reference/large language model/knowledge self-distillation/evaluation of innovation/text classification/sci-ence and technology evaluation分类
社会科学引用本文复制引用
邱昕鹏,李晶..基于大语言模型知识自蒸馏的引用意图自动识别[J].情报杂志,2025,44(6):119-126,118,9.基金项目
国家社会科学基金项目"科技论文创新质量的微观测度及应用研究"(编号:22BTQ097)研究成果. （编号:22BTQ097）
