HC3 Plus
收藏资源简介:
HC3 Plus是由中国科学院信息工程研究所创建的一个大规模数据集,专注于检测ChatGPT生成的文本。该数据集包含210,000条样本,涵盖翻译、总结和改写等多种语义不变任务。创建过程中,研究团队首先选取了多个高质量的人工标注数据集,然后使用ChatGPT生成目标文本。HC3 Plus旨在解决当前检测模型在语义不变任务上的识别难题,特别是在区分AI生成内容和人类编写内容方面的挑战。
HC3 Plus is a large-scale dataset created by the Institute of Information Engineering, Chinese Academy of Sciences, focusing on detecting ChatGPT-generated text. This dataset contains 210,000 samples, covering a variety of semantics-preserving tasks including translation, summarization, and paraphrasing. During its creation, the research team first selected multiple high-quality manually annotated datasets, then used ChatGPT to generate target texts. HC3 Plus aims to address the recognition challenges of current detection models on semantics-preserving tasks, particularly the challenge of distinguishing between AI-generated content and human-written content.

- 1HC3 Plus: A Semantic-Invariant Human ChatGPT Comparison Corpus中国科学院信息工程研究所 · 2024年



