OmniOpt: Taxonomy, Geometry, and Benchmarking of Modern Optimizers

Hugging Face Daily Papers Papers

Summary

OmniOpt presents a unified framework for optimizer selection in large-scale model training, combining meta-pipeline transformations, norm-constrained linear minimization oracles, and a cross-domain benchmark to systematically analyze optimizer families and their trade-offs.

Optimizer selection for large-scale model training has become a system-level design decision constrained jointly by compute, memory, tuning budget, and task diversity, yet the landscape of over one hundred methods remains fragmented. We therefore present OmniOpt, a unified survey and benchmark cookbook of optimizers for the research community. OmniOpt rests on four coupled components. First, we treat every optimizer update as a structured transformation through a five-stage meta-pipeline, and show that most methods engage only one or two of these stages. Second, we use norm-constrained linear minimization oracles (LMOs) to unify different optimizers. Third, these two views ground a dual-dimension taxonomy, one dimension assigning each method to a mechanism family and the other recording the measurable training objectives it aims to improve. Fourth, and at the core of this paper, we instantiate the full taxonomy in a unified cross-domain benchmark spanning representative optimizers, model scales, and training regimes from language model pretraining to image classification, systematically analyzing each method family across multiple effect objectives and laying out their trade-offs. OmniOpt thus supplies the research community with an operational coordinate system for selecting optimizers under explicit mechanism and objective assumptions, and charts a direction for the future development of the optimizer community.
Original Article
View Cached Full Text

Cached at: 07/07/26, 06:42 AM

Paper page - OmniOpt: Taxonomy, Geometry, and Benchmarking of Modern Optimizers

Source: https://huggingface.co/papers/2607.04033 Authors:

,

,

,

,

,

,

,

,

,

,

Abstract

OmniOpt presents a unified framework for optimizer selection in large-scale model training by combining meta-pipeline transformations, norm-constrained linear minimization oracles, and a cross-domain benchmark to systematically analyze optimizer families and their trade-offs across different training objectives and model scales.

Optimizer selectionforlarge-scale model traininghas become a system-level design decision constrained jointly by compute, memory, tuning budget, and task diversity, yet the landscape of over one hundred methods remains fragmented. We therefore present OmniOpt, a unified survey and benchmark cookbook of optimizers for the research community. OmniOpt rests on four coupled components. First, we treat every optimizer update as a structured transformation through a five-stagemeta-pipeline, and show that most methods engage only one or two of these stages. Second, we usenorm-constrained linear minimization oracles(LMOs) to unify different optimizers. Third, these two views ground a dual-dimension taxonomy, one dimension assigning each method to a mechanism family and the other recording the measurabletraining objectivesit aims to improve. Fourth, and at the core of this paper, we instantiate the full taxonomy in a unifiedcross-domain benchmarkspanning representative optimizers,model scales, and training regimes from language model pretraining to image classification, systematically analyzing each method family across multiple effect objectives and laying out theirtrade-offs. OmniOpt thus supplies the research community with an operational coordinate system for selecting optimizers under explicit mechanism and objective assumptions, and charts a direction for the future development of the optimizer community.

View arXiv pageView PDFProject pageGitHub5Add to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2607.04033 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.04033 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.04033 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

optimize_anything: A Universal API for Optimizing any Text Parameter

arXiv cs.CL

This paper presents optimize_anything, a universal LLM-based optimization system for text artifacts that achieves state-of-the-art results across diverse tasks including agent architecture discovery, scheduling, CUDA kernel generation, and packing, demonstrating general-purpose text optimization.

Optimistic Dual Averaging Unifies Modern Optimizers

arXiv cs.LG

This paper introduces SODA, a generalization of Optimistic Dual Averaging that unifies various modern optimizers like Muon and Lion. It proposes a practical wrapper that improves performance across different scales without requiring additional hyperparameter tuning for weight decay.