Muon Sublates the Edge of Stability in LLM Pretraining
cs.LG
Submitted: 2026-09-28
Updated: 2026-09-28
Code: https://github.com/cyzebra/Muon-Sublates-the-Edge-of-Stability
Sources
- Edge of Stochastic Stability: Revisiting the Edge of Stability for SGD
- Adaptive Gradient Methods at the Edge of Stability
- MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies
- A Scalable Measure of Loss Landscape Curvature for Analyzing the Training Dynamics of LLMs
- The large learning rate phase of deep learning: the catapult mechanism
- The Origin of Edge of Stability
- Muon is Scalable for LLM Training
- A Defense of the Quadratic Model
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