An Automated Code Review Framework Based on BERT and Qianwen Large Model
Qiang Zhu, Zhengying Wang, Yufei Zhuang, Guo Chen, He Luo, Meiqi Wu, Hongxiang Chen, Zhi Liu · 2025
As software systems become more complex and widely adopted across industries, ensuring code quality and security has become increasingly crucial. Traditional code auditing methods, such as manual review and rule-based static analysis, struggle to address the challenges posed by the scale and complexity of modern software vulnerabilities. To overcome these issues, this paper proposes an automated code auditing framework that combines BERT’s contextual understanding capabilities with the vast knowledge base of the Qianwen large language model. The framework leverages deep learning and advanced language modeling techniques to improve the efficiency, accuracy, and adaptability of vulnerability detection in complex and dynamic code environments. Experimental results show that the model achieves an accuracy of $89.05 \%$ on the Buffer Errors dataset and $\mathbf{9 7. 8 2 \%}$ on the Resource Management Errors dataset.