Jul 2026
Sharpness-Aware Minimization and Muon: Robustness under the Spectral Norm
Across ImageNet-1K experiments on ViT-Small/16 and ResNet-50, it is found that the combination of a spectral inner step with a Muon outer step performs consistently strongly, achieving the best validation accuracy on both models among the evaluated methods.
Wenzhi Zhong, Edward Milsom, Michael Murray
· arXiv.org · 0 citations