LinearizeLLM: An Agent-Based Framework for LLM-Driven Exact Linear Reformulation of Nonlinear Optimization Problems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kandora, Paul-Niklas Ken, Zeller, Simon Caspar, Elsing, Aaron Jeremias, Kuss, Elena, Rebennack, Steffen
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914300113715200
author Kandora, Paul-Niklas Ken
Zeller, Simon Caspar
Elsing, Aaron Jeremias
Kuss, Elena
Rebennack, Steffen
author_facet Kandora, Paul-Niklas Ken
Zeller, Simon Caspar
Elsing, Aaron Jeremias
Kuss, Elena
Rebennack, Steffen
contents Reformulating nonlinear optimization problems into solver-ready linear optimization problems is often necessary for practical applications, but the process is often manual and requires domain expertise. We propose LinearizeLLM, an agent-based LLM framework that produces solver-ready linear reformulations of nonlinear optimization problems. Agents first detect the nonlinearity pattern (e.g., bilinear products) and apply nonlinearity pattern-aware reformulation techniques, selecting the most suitable linearization technique. We benchmark on 40 instances: 27 derived from ComplexOR by injecting exactly-linearizable operators, and 13 automatically generated instances with deeply nested nonlinearities. LinearizeLLM achieves 73\% mean end-to-end overall success (OSR) across nonlinearity depths (8.3x higher than a one-shot LLM baseline; 4.3x higher than Pyomo). The results suggest that a set of pattern-specialized agents can automate linearization, supporting natural-language-based modeling of nonlinear optimization.
format Preprint
id arxiv_https___arxiv_org_abs_2510_15969
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LinearizeLLM: An Agent-Based Framework for LLM-Driven Exact Linear Reformulation of Nonlinear Optimization Problems
Kandora, Paul-Niklas Ken
Zeller, Simon Caspar
Elsing, Aaron Jeremias
Kuss, Elena
Rebennack, Steffen
Machine Learning
Artificial Intelligence
Reformulating nonlinear optimization problems into solver-ready linear optimization problems is often necessary for practical applications, but the process is often manual and requires domain expertise. We propose LinearizeLLM, an agent-based LLM framework that produces solver-ready linear reformulations of nonlinear optimization problems. Agents first detect the nonlinearity pattern (e.g., bilinear products) and apply nonlinearity pattern-aware reformulation techniques, selecting the most suitable linearization technique. We benchmark on 40 instances: 27 derived from ComplexOR by injecting exactly-linearizable operators, and 13 automatically generated instances with deeply nested nonlinearities. LinearizeLLM achieves 73\% mean end-to-end overall success (OSR) across nonlinearity depths (8.3x higher than a one-shot LLM baseline; 4.3x higher than Pyomo). The results suggest that a set of pattern-specialized agents can automate linearization, supporting natural-language-based modeling of nonlinear optimization.
title LinearizeLLM: An Agent-Based Framework for LLM-Driven Exact Linear Reformulation of Nonlinear Optimization Problems
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.15969