郑州大学学报（理学版）2026，Vol.58Issue(5)：45-53,9.DOI:10.13705/j.issn.1671-6841.2025045
基于大语言模型的因果分析类案检索方法
LLM-based Causal Analysis Method for Legal Case Retrieval
摘要
Abstract
Legal case retrieval was recognized as a critical component in promoting judicial fairness and consistency.Existing approaches,however,were primarily based on textual statistical features and shal-low semantics,facing limitations such as lengthy documents,redundant information,and a lack of causal logic inherent in legal reasoning.To address these challenges,a large language model(LLM)-based causal analysis method was introduced.The proposed framework operated through a three-stage process:fact reconstruction,behavioral fact extraction,and causal discovery.Case facts were reconstructed by fil-tering procedural noise;core behavioral sequences were extracted using knowledge graphs;counterfactual reasoning was used to identify causal relationships;and causal elements were inferred from behavioral facts with the assistance of LLMs,thereby uncovering deeper judicial logic.Experimental results showed that baseline models fine-tuned with causal features derived from this framework achieved substantial per-formance in indexes as P@,P@10,MAP,NDCG@10,NDCG@20,and NDCG@30.These findings validated the effectiveness of integrating causal analysis into legal case retrieval systems to enhance accu-racy and ranking quality.关键词
智慧法院/裁判文书/类案检索/大语言模型/因果分析Key words
smart court/judicial document/case law retrieval/large language model/casual analysis分类
信息技术与安全科学引用本文复制引用
王凌志,周裕林,秦永彬,黄瑞章,陈艳平,林川..基于大语言模型的因果分析类案检索方法[J].郑州大学学报（理学版）,2026,58(5):45-53,9.基金项目
国家重点研发计划项目(2023YFC3304500) （2023YFC3304500）
贵州省科学技术基金重点资助项目(黔科合重大专项字[2024]003) （黔科合重大专项字[2024]003）
贵州省研究生科研基金项目(2024YJSKYJJ041) （2024YJSKYJJ041）
黔科合支撑[2023]一般项目(448) （448）
