Navan Preet Singh, Anurag Garikipati, Ahmed Abulkhair, Jyani Akshay Jagdishbhai, Atul Yaduvanshi, Amarendra Chaudhary, Madalina Ciobanu, Qingqing Mao, Ritankar Das

State-of-the-Art Arabic Language Modeling with Sparse MoE Fine-Tuning and Chain-of-Thought Distillation

arXiv:2604.06421v1 Announce Type: new Abstract: This paper introduces Arabic-DeepSeek-R1, an application-driven open-source Arabic LLM that leverages a sparse MoE backbone to address the digital equity …