In a recent arXiv preprint, Gabriel Peyré studies an idealized deterministic, continuous-time, vanishing-momentum version of Muon in the mean-field regime, where wide models are represented by probability measures on parameter space. The paper introduces Spectral Wasserstein distances indexed by norms on positive semidefinite matrices, with the trace norm recovering the classical W2 distance and the operator norm giving the Muon geometry. Schatten norms interpolate between these extremes.

The work develops a static Kantorovich formulation, a max-min robust-cost representation, and Gaussian reductions extending the Bures formula. For monotone norms, the author proves equivalence with a Benamou-Brenier formulation, yielding a gradient-flow interpretation of normalized mean-field training dynamics. Numerical experiments illustrate the resulting flows for MMD objectives, Gaussian reductions, two-layer ReLU models, and shallow attention.

Since there is only one source, there are no conflicting claims to compare; the article is a self-contained theoretical contribution rather than a survey or debate.