The rapid rise of AI code generation tools has created significant challenges for computer science educators in verifying the authenticity of students' code writing. While generic AI detection tools exist, they often struggle to accurately identify AI-generated code because of the unique patterns and structures of programming languages. Research in this area requires data to develop effective systems for detecting AI code. We introduce a specialized dataset designed to support the creation of domain-specific detection tools to fill this gap. The dataset comprises 10,000 annotated code samples, consisting of 5000 human-written and 5000 AI-generated code samples, across Python, Java, C, and C++. The human-written samples were collected from a public repository, while the AI-generated samples were produced using ChatGPT's API with varied prompts. Each sample is labeled with its origin, whether human or AI, enabling the robust training of machine learning and deep learning models for code source discrimination. The dataset and experiment code are publicly available to support further research in AI-generated code detection.
boukili et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: