Two-sided diagonal preconditioning before Newton-Schulz orthogonalization, plus an adaptive scalar stepsize, improves Muon's GPT-2 pretraining loss at nearly unchanged cost.
Natural gradient works efficiently in learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
MALT: Lightweight Curvature-Aware Muon via Diagonal Preconditioning
Two-sided diagonal preconditioning before Newton-Schulz orthogonalization, plus an adaptive scalar stepsize, improves Muon's GPT-2 pretraining loss at nearly unchanged cost.