Pion: A Spectrum-Preserving Optimizer via Orthogonal Equivalence Transformation

Hugging Face Daily Papers Papers

Summary

This paper introduces Pion, a novel spectrum-preserving optimizer for large language model training that uses orthogonal equivalence transformations to maintain singular values during weight updates, offering stable performance comparable to standard optimizers.

We introduce Pion, a spectrum-preserving optimizer for large language model (LLM) training based on orthogonal equivalence transformation. Unlike additive optimizers such as Adam and Muon, Pion updates each weight matrix through left and right orthogonal transformations, preserving its singular values throughout training. This yields an optimization mechanism that modulates the geometry of weight matrices while keeping their spectral norm fixed. We derive the Pion update rule, systematically examine its design choices, and analyze its convergence behavior along with several key properties. Empirical results show that Pion offers a stable and competitive alternative to standard optimizers for both LLM pretraining and finetuning.
Original Article
View Cached Full Text

Cached at: 05/13/26, 12:15 PM

Paper page - Pion: A Spectrum-Preserving Optimizer via Orthogonal Equivalence Transformation

Source: https://huggingface.co/papers/2605.12492

Abstract

Pion is a spectrum-preserving optimizer for large language model training that uses orthogonal equivalence transformations to maintain singular values during weight updates, offering stable performance comparable to standard optimizers.

We introduce Pion, a spectrum-preserving optimizer for large language model (LLM) training based onorthogonal equivalence transformation. Unlikeadditive optimizerssuch asAdamandMuon, Pion updates each weight matrix through left and rightorthogonal transformations, preserving itssingular valuesthroughout training. This yields anoptimization mechanismthat modulates the geometry ofweight matriceswhile keeping theirspectral normfixed. We derive the Pion update rule, systematically examine its design choices, and analyze itsconvergence behavioralong with several key properties. Empirical results show that Pion offers a stable and competitive alternative to standard optimizers for both LLM pretraining and finetuning.

View arXiv pageView PDFProject pageGitHub3Add to collection

Get this paper in your agent:

hf papers read 2605\.12492

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.12492 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.12492 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.12492 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Anytime Training with Schedule-Free Spectral Optimization

arXiv cs.LG

This paper introduces SF-NorMuon, a schedule-free spectral optimizer that matches or exceeds tuned AdamW on language models up to 772M parameters, with theoretical guarantees for stationarity and long-horizon stability.

Muon$^p$: Muon with Fractional Spectral Powers

arXiv cs.LG

This paper introduces Muon^p, a novel optimizer that uses fractional spectral-power updates to interpolate between Muon and gradient descent, providing theoretical justification and empirical gains on billion-scale fine-tuning tasks.

Aurora: A Leverage-Aware Spectral Optimizer

arXiv cs.LG

Aurora is a leverage-aware spectral optimizer that addresses neuron death in MLP layers by enforcing row uniformity while preserving the polar factor geometry of Muon updates, achieving state-of-the-art performance on the modded-nanoGPT speedrun benchmark.