Id: arxiv/2602.11948 Url: https://arxiv.org/abs/2602.11948 Tags: math.OC, cs.LG Score: 0 Title: Insights on Muon from Simple Quadratics Domain: math.OC Source: arxiv Authors: Antoine Gonon, Andreea-Alexandra Muşat, Nicolas Boumal Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-02-12 Page: https://delphi.science/r/arxiv/2602.11948 Muon updates weight matrices along (approximate) polar factors of the gradients and has shown strong empirical performance in large-scale training. Existing attempts at explaining its performance largely focus on single-step comparisons (on quadratic proxies) and worst-case guarantees that treat the inexactness of the polar-factor as a nuisance ``to be argued away''. We show that already on simple strongly convex functions such as $L(W)=\frac12\|W\|_{\text{F}}^2$, these perspectives are insufficient, suggesting that understanding Muon requires going beyond local proxies and pessimistic…