The study introduces the xFakeSci learning algorithm to identify fake science generated by ChatGPT. It demonstrates distinctive behaviors between ChatGPT content and scientific articles. The algorithm achieves F1 scores ranging from 80% to 94%, outperforming state-of-the-art algorithms. By training on diverse datasets, including PubMed articles and ChatGPT-generated documents, xFakeSci effectively predicts the authenticity of biomedical articles. Calibration using data-driven heuristics enhances prediction accuracy, mitigating overfitting issues. The research highlights the importance of combating fake science in the era of generative AI tools like ChatGPT.
To Another Language
from source content
arxiv.org
Viktige innsikter hentet fra
by Ahmed Abdeen... klokken arxiv.org 03-14-2024
https://arxiv.org/pdf/2308.11767.pdfDypere Spørsmål