Muon Is Theoretically Wrong For Convolutions, But Empirically Effective

cs.LG, cs.AI

Submitted: 2026-10-05

Updated: 2026-10-05

Code: https://github.com/thib-s/muonconv-cifar10-airbench

Project page: https://leloykun.github.io/ponder/freqmuon/.5https://www.deel.ai/10

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• Scalable Second Order Optimization for Deep Learning
• Spectral Norm of Convolutional Layers with Circular and Zero Paddings
• Muon is Scalable for LLM Training
• Controlled LLM Training on Spectral Sphere

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks