TL;DR
ArabiDeepfake provides detailed, reproducible protocols that advance robust Arabic deepfake detection and compares a pooled single model against a five-model ensemble against a five-model ensemble, reporting 95% CIs and McNemar tests.
A high-quality Arabic dataset spanning six major dialects and seven distinct domains is compiled, providing a representative benchmark for real-world environments and suggesting that the model effectively identifies the stylistic and expressive markers inherent in synthetic conversational Arabic.
Deepfake detectors are rarely evaluated on Arabic-media content. We introduce Arab-DF, an Arabic-media visual deepfake dataset of 1066 authentic and 11,240 manipulated videos covering four manipulation families. Thirteen public detector checkpoints are evaluated zero-shot under a common protocol, and each must first re...
In this paper, we introduce MAQASID, a closed-domain dataset for exploring the optimal semantic representation of Indonesian Islamic texts. It is designed to solve challenges in low-resource languages and closed-domain religious datasets, such as Arabic-to-Indonesian transliteration inconsistencies and semantic shifts....
Per-class analysis shows every model struggles with several minority categories, indicating Arabic dark pattern detection remains genuinely difficult at current data volumes, indicating pre-training on dialectal, code-switched Arabic may matter more here than Arabic pre-training alone.
This work offers a robust benchmark dataset, a three-tier labeling framework, and insights into preprocessing strategies tailored to Arabic’s linguistic complexity, advancing the development of culturally aligned and ethically sound Arabic large language models for content moderation and beyond.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
