Shulou(Shulou.com)11/24 Report--

CTOnews.com, June 8 (Xinhua)-- according to Japan's Asahi Shimbun, a research team at the University of Kansas recently found a new test that can accurately determine whether a paper was written by ChatGPT with an accuracy of 99%.

At present, users only need to enter a set of data and say, "Please use these data to write an academic paper." ChatGPT can create a fake paper in tens of seconds, which makes many authors already use ChatGPT to write their papers.

However, some academic journals and research institutions have warned that plagiarism or other fraud would be committed if the authors did not explicitly state that they used ChatGPT at the time of submission.

For this reason, progress has been made in the development of tools for testing whether articles are generated by AI, but so far, there is no testing tool specifically for academic papers.

The University of Kansas team focused on "paragraph complexity" and "common phrases" to train a new AI to test whether a paper was generated by ChatGPT.

In the six months to March 2023, the research team selected 64 papers from the US academic journal Science and asked ChatGPT to generate abstracts of about 300,400 words, resulting in 128 articles written by ChatGPT and used to train the new test AI model, the report said.

After training, the model can identify the artificially created parts from the articles generated by AI. The recognition accuracy of specific paragraphs in the text is 92%, and the comprehensive accuracy is 99%. The results have been published in Cell Reports Physical Science magazine.

The research team found that ChatGPT rarely uses expressions such as "although... but (but, however, although)". On the contrary, it likes to use words such as "others" and "researchers".

On the other hand, human beings are more accustomed to the interlaced use of long sentences and short sentences, and their sentence patterns are different from those of ChatGPT.

CTOnews.com previously reported that a professor at Texas A & M University in the United States had inversely verified whether they were also generated by ChatGPT by typing student papers into ChatGPT, which also achieved "good results"-directly causing students in half the class to fail.

CTOnews.com attached a link to the paper: click here to

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.