CodeSmellData
收藏资源简介:
CodeSmellData是由威廉与玛丽学院和达卡大学的研究团队创建的数据集,旨在评估大型语言模型生成代码异味的倾向性。该数据集包含142,817条方法级代码异味实例,涵盖了13种不同类型的代码异味,数据来源于GitHub上高星级的Python开源项目。数据集通过Pylint工具分析并手动验证,确保了数据的准确性和代表性。CodeSmellData的应用领域主要集中在软件工程中代码质量的评估,特别是通过分析大型语言模型生成的代码中可能存在的设计缺陷,帮助开发者更好地理解和改进代码生成模型。
CodeSmellData is a dataset created by research teams from the College of William & Mary and the University of Dhaka, aiming to evaluate the propensity of large language models (LLMs) to generate code smells. This dataset contains 142,817 method-level code smell instances covering 13 distinct types of code smells, and is sourced from high-starred Python open-source projects on GitHub. The dataset was analyzed via the Pylint tool and manually verified to ensure its accuracy and representativeness. The main application domain of CodeSmellData focuses on code quality assessment in software engineering, specifically by analyzing potential design defects in code generated by large language models to help developers better understand and improve code generation models.

- 1How Propense Are Large Language Models at Producing Code Smells? A Benchmarking Study威廉与玛丽学院, 达卡大学 · 2024年



