Preprint
Aug 2026
Dion3: Full-Stack Orthogonal Updates
Dion3, a revision of Muon that targets this overhead at every level of the stack with a simple change to the update rule that cuts costs even further: selecting only a fraction of the momentum matrix's rows to orthogonalize at each step.
Noah Amsel, Jack Zhang, Kwangjun Ahn et al.
· 0 citations