This dataset offers machine generated and human generated texts split over various tasks, languages, generation models, and prompting styles.
AITextBench is a dataset for machine generated text detection. It consists of 28967 texts with labels. The machine generated text came from four different genration models: Llama-3-8b, Phi-3-Mini, Mistral, and GPT-4o. There are three different prompting styles used to get the… See the full description on the dataset page:
https://huggingface.co/datasets/btufts/AITextBench-V2.