AIGCodeSet: A New Annotated Dataset for AI Generated Code Detection

TL;DR

This paper introduces AIGCodeSet, a new annotated dataset of human and AI-generated Python code, and evaluates baseline detection methods, highlighting the Bayesian classifier's superior performance.

Contribution

The creation of AIGCodeSet, a large, annotated dataset for AI-generated code detection, and the experimental comparison of baseline detection methods.

Findings

Bayesian classifier outperforms other models in detection accuracy.

AIGCodeSet contains 2,828 AI-generated and 4,755 human-written Python codes.

Experiments demonstrate effectiveness of baseline detection methods.

Abstract

While large language models provide significant convenience for software development, they can lead to ethical issues in job interviews and student assignments. Therefore, determining whether a piece of code is written by a human or generated by an artificial intelligence (AI) model is a critical issue. In this study, we present AIGCodeSet, which consists of 2.828 AI-generated and 4.755 human-written Python codes, created using CodeLlama 34B, Codestral 22B, and Gemini 1.5 Flash. In addition, we share the results of our experiments conducted with baseline detection methods. Our experiments show that a Bayesian classifier outperforms the other models.