Please use this identifier to cite or link to this item: https://dspace.chmnu.edu.ua/jspui/handle/123456789/3416
Title: A Metric-Driven Evaluation of Rephrased and Generated Texts
Authors: Antipova, K.
Horban, H.
Keywords: academic abstracts
ai-generated texts
detectors
large language models
natural language processing
stylometric analysis
Issue Date: 2025
Publisher: CEUR-WSc/o Manfred Jeusfel (Sweden)
Abstract: The rapid development of large language models has raised serious concerns about the reliability of detecting content created by artificial intelligence. This article compares the stylistic metrics of texts generated using a multimodal model and an autoregressive model. The results show that the generated text is very similar to human-written text in terms of lexical diversity and semantic coherence. In terms of perplexity and burstiness, the rephrased texts are practically indistinguishable from the original human-written texts, which leads to a high level of false negatives in autoregressive detectors. Our analysis highlights the need for new detection methods and suggests further directions, including more specific stylometric signatures. Relying solely on a single stylometric metric leads to unreliable differentiation between generated and human-written text.
Description: Antipova, K., & Horban, H. (2025). A Metric-Driven Evaluation of Rephrased and Generated Texts In : V. Snytyuk, L. Kirichenko, O. Mulesa, S. Lupenko (eds.). Artificial Intelligence Technologies and Data Science 2025, IT&I-WS: AITDS 2025 : 12th Information Technology and Implementation Workshop-IT&I-WS : Proceedings Paper, 4158, (51–60). CEUR-WSc/o Manfred Jeusfeld. Skovde.
URI: https://www.webofscience.com/wos/woscc/full-record/WOS:001902110900005
https://dspace.chmnu.edu.ua/jspui/handle/123456789/3416
ISSN: 1613-0073
Appears in Collections:Публікації науково-педагогічних працівників ЧНУ імені Петра Могили у БД Web of Science

Files in This Item:
File Description SizeFormat 
Antipova, K., Horban, H.pdf80.6 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.