English
Russian
French
Spanish
German
Japanese
korean
PortugueseNews
Oct 10, 2025
Optimizing LLM Pre-Training: Muon, Latent Attention, and MoE in Practice
Muon is a geometry-aware optimizer that halves training time for large language models. It uses polar decomposition and spectral n...
Muon is a geometry-aware optimizer that halves training time for large language models. It uses polar decomposition and spectral n...
