Dmitrii Vlashchuk

Papers that credit Dmitrii Vlashchuk in their source bylines. Matching names may refer to different researchers.

1 results

1. PaperMachine Learning

   Structured weights make Muon training cheaper at scale

   In matched dense and MoE runs, Monarch–Muon had lower loss than Monarch–AdamW; at 6.89B, it cut step time and memory versus dense Muon.

   Added to Chiosa Oct 14, 2026