This model detects whether text is AI generated vs created by a human and outputs the percentage liklihood that the text is generated by AI (From smaller LLM’s like LLAMA to ChatGPT, GPT-4, etc)

This model is using the methods from the paper Fast-DetectGPT: Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvatureere from ICLR 2024.

@misc{bao2024fastdetectgpt, title={Fast-DetectGPT: Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature}, author={Guangsheng Bao and Yanbin Zhao and Zhiyang Teng and Linyi Yang and Yue Zhang}, year={2024}, eprint={2310.05130}, archivePrefix={arXiv}, primaryClass={cs.CL} }

Note that it works better the longer the text is and it is slow to boot up but fast once it’s warmed up.