We examine the impact of homograph attacks on the Sentiment Analysis (SA) task of different Arabic dialects from the Maghreb North-African countries. Homograph attacks result in a 65.3% decrease in transformer classification from an F1-score of 0.95 to 0.33 when data is written in "Arabizi". The goal of this study is to highlight LLMs weaknesses' and to prioritize ethical and responsible Machine Learning.

翻译：我们研究了同形词攻击对来自马格里布北非国家的不同阿拉伯方言情感分析任务的影响。当数据以"Arabizi"书写时，同形词攻击导致Transformer分类的F1分数从0.95降至0.33，下降幅度达65.3%。本研究的目的是揭示大语言模型的弱点，并优先考虑道德和负责任的机器学习。