Work

PTUK-HULAT at AraGenEval Shared Task: Fine-tuning XLM-RoBERTa for AI-Generated Arabic News Detection

Year: 2025

Type: conference-paper

Abstract: The authenticity of digital content has become an increasingly critical challenge with the rapid adoption of generative AI tools, especially for low-resource languages such as Arabic.The language's rich morphology and domain diversity further complicate the detection of machine-generated Arabic text.In this work, we present our submission to the ARATECT 4.3 shared task, Subtask 3, which focuses on identifying AI-generated Arabic news articles.Our approach employs fine-tuned multilingual transformer models based on XLM-RoBERTa.The XLM-RoBERTa-large model achieved a macro F1-score of 0.93 on the development set, while the XLM-RoBERTa-base model obtained an F1-score of 0.78 on the test set, ranking fourth on the official leaderboard.This paper outlines our methodology, presents the experimental results, and discusses key insights from our participation.

Source: Arabic Natural Language Processing Workshop (WANLP)

DOI: 10.18653/v1/2025.arabicnlp-sharedtasks.7

Authors: Tasneem Duridi, Areej Jaber, Paloma Martínez

Institutions: Palestine Technical University - Kadoorie, Universidad Carlos III de Madrid

Language: English

Topic: Authorship Attribution and Profiling

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: Palestine Technical University Kadoorie, Universidad Carlos III de Madrid

OpenAlex ID: W4416037833