# Demystifying Manifold Constraints in LLM Pre-training

arXiv: https://arxiv.org/abs/2605.04418 · HTML: https://mlsanity.com/paper/2605.04418/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 56 of its 80 references that are in the map.
- First posted on arXiv 6 May 2026.

## Builds on (most cited first, 30 of 56)

- [Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift](https://mlsanity.com/paper/1502.03167/) (2015)
- [Layer Normalization](https://mlsanity.com/paper/1607.06450/) (2016)
- [Spectral Normalization for Generative Adversarial Networks](https://mlsanity.com/paper/1802.05957/) (2018)
- [Group Normalization](https://mlsanity.com/paper/1803.08494/) (2018)
- [Root Mean Square Layer Normalization](https://mlsanity.com/paper/1910.07467/) (2019)
- [Scaling Language Models: Methods, Analysis & Insights from Training Gopher](https://mlsanity.com/paper/2112.11446/) (2021)
- [Mixed Precision Training](https://mlsanity.com/paper/1710.03740/) (2017)
- [Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks](https://mlsanity.com/paper/1602.07868/) (2016)
- [Shampoo: Preconditioned Stochastic Tensor Optimization](https://mlsanity.com/paper/1802.09568/) (2018)
- [Analyzing and Improving the Training Dynamics of Diffusion Models](https://mlsanity.com/paper/2312.02696/) (2023)
- [Deep Learning with Limited Numerical Precision](https://mlsanity.com/paper/1502.02551/) (2015)
- [Lower Bounds for Non-Convex Stochastic Optimization](https://mlsanity.com/paper/1912.02365/) (2019)
- [Old Optimizer, New Norm: An Anthology](https://mlsanity.com/paper/2409.20325/) (2024)
- [Training Deep Learning Models with Norm-Constrained LMOs](https://mlsanity.com/paper/2502.07529/) (2025)
- [BiLLM: Pushing the Limit of Post-Training Quantization for LLMs](https://mlsanity.com/paper/2402.04291/) (2024)
- [L2 Regularization versus Batch and Weight Normalization](https://mlsanity.com/paper/1706.05350/) (2017)
- [An Exponential Learning Rate Schedule for Deep Learning](https://mlsanity.com/paper/1910.07454/) (2019)
- [Three Mechanisms of Weight Decay Regularization](https://mlsanity.com/paper/1810.12281/) (2018)
- [A Spectral Condition for Feature Learning](https://mlsanity.com/paper/2310.17813/) (2023)
- [Muon Optimizes Under Spectral Norm Constraints](https://mlsanity.com/paper/2506.15054/) (2025)
- [TorchTitan: One-stop PyTorch native solution for production ready LLM pre-training](https://mlsanity.com/paper/2410.06511/) (2024)
- [A Distributed Data-Parallel PyTorch Implementation of the Distributed Shampoo Optimizer for Training Neural Networks At-Scale](https://mlsanity.com/paper/2309.06497/) (2023)
- [Why Do We Need Weight Decay in Modern Deep Learning?](https://mlsanity.com/paper/2310.04415/) (2023)
- [nGPT: Normalized Transformer with Representation Learning on the Hypersphere](https://mlsanity.com/paper/2410.01131/) (2024)
- [Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate](https://mlsanity.com/paper/2010.02916/) (2020)
- [Implicit Bias of AdamW: ℓ_∞ Norm Constrained Optimization](https://mlsanity.com/paper/2404.04454/) (2024)
- [Rotational Equilibrium: How Weight Decay Balances Learning Across Neural Networks](https://mlsanity.com/paper/2305.17212/) (2023)
- [Lion Secretly Solves Constrained Optimization: As Lyapunov Predicts](https://mlsanity.com/paper/2310.05898/) (2023)
- [Understanding AdamW through Proximal Methods and Scale-Freeness](https://mlsanity.com/paper/2202.00089/) (2022)
- [Decoupled Networks](https://mlsanity.com/paper/1804.08071/) (2018)

## Built on by (most cited first, 1 of 1)

- [Curvature-Conditioned Multiscale Momentum with Sphere Constraints for LLM Pretraining](https://mlsanity.com/paper/2608.28442/) (2026)

## Abstract (arXiv)

The recent success of matrix optimizers (e.g., Muon) suggests that specific normalization of momentum, such as orthogonalization and row-wise normalization, benefits both the stability and acceleration of LLM training. Consequently, several recent studies have suggested that weights should also be normalized, leading to a Riemannian optimization problem. While such constrained training frameworks demonstrate superior performance, the effects of explicitly constraining weights, and their interaction with existing stabilization mechanisms, remain less understood. To bridge this gap, we study manifold constrained training dynamics through activation scales, rotational dynamics, and the update-to-weight ratio. We propose a Riemannian spectral steepest descent optimizer called MACRO, alongside a radius selection principle to serve as our testbed. Our analysis and numerical experiments reveal that RMSNorm and manifold constraints serve overlapping roles, and that weight decay can be completely eliminated when manifold constraints are applied. By controlling the update-to-weight ratio, constrained training significantly alleviates update cancellation, empirically demonstrating that MACRO is robust to low-precision computation and competitive with existing algorithms for standard LLM pre-training.