Year: 2025
Type: conference-paper
Abstract: This paper describes the approach developed for the AraGenEval shared task, with a focus on Arabic authorship identification and AI-generated text detection.Transformerbased models, including ALLaM-7B-Instructpreview for Subtask 2 and AraModernBERT for Subtask 3, were fine-tuned using both the official and additional datasets.Prompt engineering and transfer learning techniques were adapted to address challenges specific to the Arabic language.Competitive performance was achieved on both subtasks, and all code and resources have been made publicly available to facilitate reproducibility.
Author: Thamer Maseer Alharbi
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4416037842