Inner Momentum for Differentially Private Muon

arXiv:2610.02738 · cs.LG, cs.CR · Submitted 2026-10-02 · Read on arXiv

cs.LG, cs.CR

Submitted: 2026-10-02

Updated: 2026-10-02

Comments: LA-UR Number: LA-UR-26-28799

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

Terminology

• Differentially Private — 1× in this paper · explained in 3 episodes

Sources

• Automatic Clipping: Differentially Private Deep Learning Made Easier and Stronger
• DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
• Convergence of Muon with Newton-Schulz
• DP-Muon: Differentially Private Optimization via Matrix-Orthogonalized Momentum
• Kimi K2: Open Agentic Intelligence
• Muon is Scalable for LLM Training

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