Meta

Latest AI news, models and releases from Meta. ['AI assistant', 'Astryx', 'Bonsai-27B', 'Business Agent', 'BYT5', 'CodeLlama', 'expressive voice AI', 'FAIRChem v2 UMA', 'Gemma 4', 'Gemma4:e2b', 'HuBERT', 'llama', 'Llama', 'LLaMA', 'Llama-2 70B', 'LLaMA2-chat-70B', 'Llama 3', 'Llama 3.1', 'Llama 3.1-405B', 'Llama-3.1-405B', 'LLaMA-3.1 405B', 'Llama 3.1-70B', 'Llama-3.1-70B', 'Llama 3.1 8B', 'Llama 3.1-8B', 'Llama-3.1-8B', 'Llama 3.2', 'LLaMA 3.2', 'Llama-3.2-1B-Instruct', 'Llama 3.3', 'llama-3.3-70b', 'Llama 3.3-70B', 'Llama 3 8B', 'Llama3-8B-Instruct', 'Llama 4', 'Llama 4 Maverick', 'LLaMA 65B', 'Llama 70B', 'LLaVA', 'M2M100']

Official site ↗

Research 🇷🇺

Generating Synthetic Data for Mixed Russian-Kazakh Code-Switching Translation

Researchers from MWS AI and several universities developed a method to generate synthetic parallel corpora for translating Russian-Kazakh code-switching (mix of languages). Using SimAlign-based word replacement, they fine-tuned models like NLLB-3.3B, achieving a human-evaluated score of 3.09/5, outperforming commercial systems (2.80-3.49) despite synthetic data limitations.

Jensen Huang's Coalition of 35 Companies Against Frontier Labs

NVIDIA CEO Jensen Huang published his first X post with an open letter 'Open Weights and American AI Leadership,' signed by 35 companies including NVIDIA, Microsoft, Meta, and others, but notably not Anthropic, Google, or Amazon. The letter defends open-weight models and distillation, amid allegations of Chinese AI model distillation from US frontier models.