project-droid/DroidCollection

该数据集是一个包含代码样本的数据集，旨在研究代码生成和检测。它分为四类主要代码：人类编写的代码、AI生成的代码、机器精炼的代码（人类和AI合作的产物）和AI生成的对抗性代码。数据来源于通用代码、算法问题代码和研究代码三个领域，并使用了11个不同的AI模型家族生成代码。数据集还包括了多种生成方法和人类与AI合作场景的模拟。

The dataset is a collection of code samples designed to study code generation and detection. It is divided into four primary categories: human-written code, AI-generated code, machine-refined code (a collaboration between humans and AI), and AI-generated adversarial code. The data sources cover three domains: general use code, algorithmic problem code, and research code, and it utilizes 11 different AI model families for code generation. The dataset also includes simulations of various generation methods and human-AI collaboration scenarios.