We’re excited to launch MAHED 2025, the first multimodal shared task on Hope and Hate Detection in Arabic content, part of ArabicNLP 2025 @ EMNLP 2025. This challenge promotes ethical AI in Arabic by addressing hate speech, hope speech, and emotion detection in both text and memes.

Website: https://marsadlab.github.io/mahed2025/

Three Tasks:

1. Hope & Hate Speech Classification Classify Arabic text as hope, hate, or neutral. Register: https://www.codabench.org/competitions/9136/ 2. Multitask Emotion & Offensive Detection Detect emotions, offensive language, and hate in text. Register: https://www.codabench.org/competitions/9166/ 3. Multimodal Hateful Meme Detection Identify hate in Arabic memes (text + image). Register: https://www.codabench.org/competitions/9192/

Key Dates:

* June 10: Data & scripts released * July 20: Final registration & test set * July 25: Submission deadline * Nov 5–9: Workshop @ EMNLP 2025, Suzhou, China