Mohamed Bayan Kmainasi retweeted
We are pleased to announce ArGuard 2026: Harmful Content Detection in Arabic Memes and LLM Prompts, a shared task co-organized by MARSAD Lab (marsadlab.com) as part of the @QRDI funded NPRP 14 Cluster Grant (NPRP 14C-0916-210015).
As Arabic digital spaces continue to grow, there is an increasing need for robust methods to identify harmful content that is often nuanced, culturally embedded, and expressed across multiple modalities. ArGuard addresses these challenges through two competitive tracks:
๐น Track A: Multimodal Hateful Meme Detection
๐น Track B: Harmful Prompt Detection for LLMs
The shared task focuses on Arabic-specific challenges, including:
โข Dialectal variation
โข Sarcasm and implicit meaning
โข Cultural references
โข Code-switching
โข Context-dependent harmful intent
๐ Website: araieval.github.io/ArGuard20โฆ
๐ Data & Code: github.com/araieval/ArGuard-โฆ
Important Dates
๐
August 1, 2026 โ Registration deadline & test set release
๐
August 6, 2026 โ Evaluation deadline
๐
August 8, 2026 โ Leaderboard release
๐
August 18, 2026 โ Shared task papers due
๐
August 30, 2026 โ Notification of acceptance
๐
September 10, 2026 โ Camera-ready papers due
๐
October 24โ29, 2026 โ ArabicNLP 2026 Conference
We welcome participation from researchers, practitioners, and students working in NLP, multimodal AI, content moderation, and LLM safety.
#ArabicNLP #EMNLP2026 #ArabicAI #NLP #LLMSafety #MultimodalAI #ResponsibleAI #ContentModeration #ArabicLanguageTechnology #AIResearch @NUQatar @NorthwesternU @HBKU @qrdicouncil


