Authorship Verification Using Cloze Test with Large Language Models

Tomáš Foltýnek, Tomáš Kancko, Pavel Rychly


Abstract
Assignment outsourcing, also known as contract cheating, occurs when a student outsources an assessment task or a part of it to a third party. It has been one of the most pressing ethical issues in university education and was further exacerbated by the wide availability of chatbots based on large language models. We propose a method that has the potential to verify the authorship of a document in question by filling in a cloze test. A close test with 10 items selected by our method can be used as a classifier with an accuracy of 0.988 and a F1 score of 0.937. We also describe a general method for building a cloze-test-based classifier when the probability of authors and non-authors correctly filling in cloze items is known.
Anthology ID:
2025.ranlp-1.45
Volume:
Proceedings of the 15th International Conference on Recent Advances in Natural Language Processing - Natural Language Processing in the Generative AI Era
Month:
September
Year:
2025
Address:
Varna, Bulgaria
Editors:
Galia Angelova, Maria Kunilovskaya, Marie Escribe, Ruslan Mitkov
Venue:
RANLP
SIG:
Publisher:
INCOMA Ltd., Shoumen, Bulgaria
Note:
Pages:
369–377
Language:
URL:
https://aclanthology.org/2025.ranlp-1.45/
DOI:
Bibkey:
Cite (ACL):
Tomáš Foltýnek, Tomáš Kancko, and Pavel Rychly. 2025. Authorship Verification Using Cloze Test with Large Language Models. In Proceedings of the 15th International Conference on Recent Advances in Natural Language Processing - Natural Language Processing in the Generative AI Era, pages 369–377, Varna, Bulgaria. INCOMA Ltd., Shoumen, Bulgaria.
Cite (Informal):
Authorship Verification Using Cloze Test with Large Language Models (Foltýnek et al., RANLP 2025)
Copy Citation:
PDF:
https://aclanthology.org/2025.ranlp-1.45.pdf