国产开源代码缺陷识别修复系统设计

打开文本图片集

中图分类号：TP311.52 文献标志码：A 文章编码：1672-7274（2025）09-0039-03

Design of a Domestic Open-source Code Defect Recognition and Repair System

AN Cheng'an

Ancheng'anE-commerceCommercial Bank，MeilanDistrict，Haikou57o1oo，Chin：

Abstract： In response to the rapid development needs of China's software industry， this article designs a domestically produced open-source code defect recognition and repair system based ona large-scale language model to repair defects indomesticsoftwarecode.The system adopts athree-tierarchitecture desig，including PC，server，and LLM server，toachievethe full processofcode defect detection，transmision，procesing，andrepair.By introducing knowledge guided open source code defect recognition methods，this study conducted experimental analysis on multipledomestic softwareand verified the efectiveness of the system in identifying and fixing code defects.The experimental results show that the new knowledge base constructed using different online LLM conversion strategies has an accuracy and recall rate of over 75% in code defect recognition and analysis， providing support for improving the quality of domestic software code.

Keywords： large scale language models; knowledge guidance; domestic software security; open source

大规模语言模型（LLM）是基于人工智能技术训练的自然语言处理模型，具有广泛的应用前景。（剩余4882字）