News Articles Dataset

该数据集包含了来自包括ABC新闻、阿尔及利亚新闻、美国新闻、美联社新闻、CBS新闻、CNN、NBC新闻、路透社和卫报等可信新闻来源的20万篇新闻文章。这些文章既有由人类撰写的，也有由大型语言模型生成的对应版本。此外，该数据集还包含了采用三种策略生成的四个大型语言模型的文章，旨在通过不同内容质量的变化帮助研究人员研究人工智能生成的假新闻的特征。数据集规模为20万篇文章，五个主题均衡分布（每个主题500篇文章）。该数据集的任务是区分人类撰写与大型语言模型生成的新闻文章。

This dataset contains 200,000 news articles sourced from reputable news outlets including ABC News, Algeria News, USA News, Associated Press (AP) News, CBS News, CNN, NBC News, Reuters, and The Guardian. These articles include both human-written originals and their corresponding versions generated by large language models (LLMs). Additionally, the dataset encompasses articles from four large language models generated via three distinct strategies, which is designed to help researchers investigate the characteristics of AI-generated fake news by introducing variations in content quality. The dataset features a balanced distribution across five topics, with 500 articles per topic, totaling 200,000 articles. The core task of this dataset is to classify news articles as either human-authored or LLM-generated.