The dataset for this challenge consists of Arabic news articles sourced from major news platforms. The goal is to build a multi-task NLP model that can analyse Arabic text and predict six target dimensions simultaneously for each article:

• Qatar Related: A binary target indicating whether the article is genuinely centered on Qatar (1) or is an incidental mention / non-Qatari event (0).
• Geography: The geographic region associated with the article (G-1 through G-12, or NO_GEOGRAPHY if not applicable).
• Politics & Conflict: The political dimension of the content (PC-1,PC-2, or NO_POLITICS).
• Health & Wellbeing: The health-related dimension of the content (H-1,H-2, or NO_HEALTH).
• Science: The scientific and technological dimension of the content (SC-1,SC-2,SC-3, or NO_SCIENCE).
• Sports: The sporting dimension of the content (SP-1 through SP-4, or NO_SPORTS).

All categorical labels have been anonymized into opaque identifiers (G-*, PC-*, H-*, SC-*, SP-*), while the explicit absence of a specific domain is represented by NO_*. The Qatar Related target is provided as a binary label (0 or 1).

Your model must learn the semantic patterns and representations directly from the Arabic text and training annotations.

Your model must learn the semantic patterns and representations directly from the Arabic text and training annotations.