Updates:
** 2023/01/18, the HC3 datasets (English & Chinese) are available at 🤗 HuggingFace & ModelScope!
** 2023/01/18, HC3 数据集 (中文版 & 英文版) 已经上线 🤗 HuggingFace & ModelScope!
HC3 (Engllish) @HuggingFace :
https://huggingface.co/datasets/Hello-SimpleAI/HC3
HC3 (Chinese) @HuggingFace :
https://huggingface.co/datasets/Hello-SimpleAI/HC3-Chinese
HC3 (英文版) @ModelScope :
https://www.modelscope.cn/datasets/simpleai/HC3
HC3 (中文版) @ModelScope :
https://www.modelscope.cn/datasets/simpleai/HC3-Chinese
We release the training and testing corpus used in our ChatGPT detector(s), including:
- 不过滤-全文 contains question-answer pairs, where the answer contains all the text without filtering indicating words. Both English and Chinese verions are provided.
- 不过滤-句子 contains question-answer pairs, where the answer contains only one splited sentence without filtering indicating words. Both English and Chinese verions are provided.
- 过滤-全文 contains question-answer pairs, where the answer contains all the text and filters the indicating words. Both English and Chinese verions are provided.
- 过滤-句子 contains question-answer pairs, where the answer contains only one splited sentence and filters the indicating words. Both English and Chinese verions are provided.
Besides, we provide DEMO for removing the indicating words in Human or ChatGPT answers, including both English and Chinese corpus.