SRLCG: Self-Rectified Large-Scale Code Generation with Multidimensional Chain-of-Thought and Dynamic Backtracking

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Ma, Hongru, Liang, Yanjie, Si, Jiasheng, Zhang, Weiyu, Guan, Hongjiao, Zheng, Chaoqun, Xu, Bing, Lu, Wenpeng
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917972949336064
author Ma, Hongru
Liang, Yanjie
Si, Jiasheng
Zhang, Weiyu
Guan, Hongjiao
Zheng, Chaoqun
Xu, Bing
Lu, Wenpeng
author_facet Ma, Hongru
Liang, Yanjie
Si, Jiasheng
Zhang, Weiyu
Guan, Hongjiao
Zheng, Chaoqun
Xu, Bing
Lu, Wenpeng
contents Large language models (LLMs) have revolutionized code generation, significantly enhancing developer productivity. However, for a vast number of users with minimal coding knowledge, LLMs provide little support, as they primarily generate isolated code snippets rather than complete, large-scale project code. Without coding expertise, these users struggle to interpret, modify, and iteratively refine the outputs of LLMs, making it impossible to assemble a complete project. To address this issue, we propose Self-Rectified Large-Scale Code Generator (SRLCG), a framework that generates complete multi-file project code from a single prompt. SRLCG employs a novel multidimensional chain-of-thought (CoT) and self-rectification to guide LLMs in generating correct and robust code files, then integrates them into a complete and coherent project using our proposed dynamic backtracking algorithm. Experimental results show that SRLCG generates code 15x longer than DeepSeek-V3, 16x longer than GPT-4, and at least 10x longer than other leading CoT-based baselines. Furthermore, they confirm its improved correctness, robustness, and performance compared to baselines in large-scale code generation.
format Preprint
id arxiv_https___arxiv_org_abs_2504_00532
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SRLCG: Self-Rectified Large-Scale Code Generation with Multidimensional Chain-of-Thought and Dynamic Backtracking
Ma, Hongru
Liang, Yanjie
Si, Jiasheng
Zhang, Weiyu
Guan, Hongjiao
Zheng, Chaoqun
Xu, Bing
Lu, Wenpeng
Software Engineering
Computation and Language
Large language models (LLMs) have revolutionized code generation, significantly enhancing developer productivity. However, for a vast number of users with minimal coding knowledge, LLMs provide little support, as they primarily generate isolated code snippets rather than complete, large-scale project code. Without coding expertise, these users struggle to interpret, modify, and iteratively refine the outputs of LLMs, making it impossible to assemble a complete project. To address this issue, we propose Self-Rectified Large-Scale Code Generator (SRLCG), a framework that generates complete multi-file project code from a single prompt. SRLCG employs a novel multidimensional chain-of-thought (CoT) and self-rectification to guide LLMs in generating correct and robust code files, then integrates them into a complete and coherent project using our proposed dynamic backtracking algorithm. Experimental results show that SRLCG generates code 15x longer than DeepSeek-V3, 16x longer than GPT-4, and at least 10x longer than other leading CoT-based baselines. Furthermore, they confirm its improved correctness, robustness, and performance compared to baselines in large-scale code generation.
title SRLCG: Self-Rectified Large-Scale Code Generation with Multidimensional Chain-of-Thought and Dynamic Backtracking
topic Software Engineering
Computation and Language
url https://arxiv.org/abs/2504.00532