Introducing the Open Arabic LLM Leaderboard

Hugging Face Blog · 2 years ago

A leaderboard platform launched to evaluate Arabic language models using standardized benchmarks has been created to address the scarcity of Arabic NLP evaluation resources. The Open Arabic LLM Leaderboard integrates 22 native Arabic datasets plus translated benchmarks like MMLU and EXAMS, serving approximately 380 million Arabic speakers. The platform enables researchers to benchmark and improve Arabic language models while plans include expanding to retrieval-augmented generation evaluation and a chatbot arena ranked by user preference.