The College of Computer Science and Information Technology organized a scientific lecture titled "Detecting Arabic Fake News Using Large Language Models."
Presented by Zahraa Mustafa Kadhum, a student at the College of Computer Science and Information Technology at University of Basrah, the lecture aimed to address the detection of fake news in Arabic by utilizing large language models, random sampling techniques, and Bayesian optimization.
The study involved constructing the "AraFND" dataset—comprising 740,465 articles—and evaluating the AraBERT-Base, MarBERT, and CAMeLBERT models. Results demonstrated the superiority of the MarBERT model, which achieved an accuracy of 92.65% at a sampling rate of 0.3. The model maintained high performance while reducing computational costs—even with limited training data—thereby supporting the advancement of Arabic language processing technologies and efforts to combat misinformation.
