AIGCodeSet
收藏资源简介:
AIGCodeSet是一个用于检测AI生成代码的数据集,由TOBB大学经济与技术大学和卡塔尔大学共同创建。该数据集专注于Python编程语言,包含4755条人类编写的代码和2828条AI生成的代码,涵盖317个编程任务。数据集的创建过程包括从CodeNet数据集中获取问题描述和人类编写的代码,使用CodeLlama 34B、Codestral 22B和Gemini 1.5 Flash模型生成AI代码,并进行后处理以确保代码质量。AIGCodeSet主要用于研究AI生成代码的检测,旨在解决教育领域中AI生成代码的伦理和学术诚信问题。
AIGCodeSet is a dataset for detecting AI-generated code, jointly created by TOBB University of Economics and Technology and Qatar University. This dataset focuses on the Python programming language, containing 4755 human-written code snippets and 2828 AI-generated code snippets, covering 317 programming tasks. The dataset creation process involves acquiring problem descriptions and human-written code from the CodeNet dataset, generating AI code using models including CodeLlama 34B, Codestral 22B, and Gemini 1.5 Flash, followed by post-processing to ensure code quality. AIGCodeSet is primarily used for research on AI-generated code detection, aiming to address ethical and academic integrity issues related to AI-generated code in the educational context.

- 1AIGCodeSet: A New Annotated Dataset for AI Generated Code DetectionTOBB大学经济与技术大学 · 2024年



