Saved in:
Bibliographic Details
Main Authors: Fang, Zhanhong, Wang, Debing, Chen, Jinbiao, Wang, Jiahai, Zhang, Zizhen
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.10148
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912706575990784
author Fang, Zhanhong
Wang, Debing
Chen, Jinbiao
Wang, Jiahai
Zhang, Zizhen
author_facet Fang, Zhanhong
Wang, Debing
Chen, Jinbiao
Wang, Jiahai
Zhang, Zizhen
contents Neural solvers have demonstrated remarkable success in combinatorial optimization, often surpassing traditional heuristics in speed, solution quality, and generalization. However, their efficacy deteriorates significantly when confronted with complex constraints that cannot be effectively managed through simple masking mechanisms. To address this limitation, we introduce Universal Constrained Preference Optimization (UCPO), a novel plug-and-play framework that seamlessly integrates preference learning into existing neural solvers via a specially designed loss function, without requiring architectural modifications. UCPO embeds constraint satisfaction directly into a preference-based objective, eliminating the need for meticulous hyperparameter tuning. Leveraging a lightweight warm-start fine-tuning protocol, UCPO enables pre-trained models to consistently produce near-optimal, feasible solutions on challenging constraint-laden tasks, achieving exceptional performance with as little as 1\% of the original training budget.
format Preprint
id arxiv_https___arxiv_org_abs_2511_10148
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle UCPO: A Universal Constrained Combinatorial Optimization Method via Preference Optimization
Fang, Zhanhong
Wang, Debing
Chen, Jinbiao
Wang, Jiahai
Zhang, Zizhen
Neural and Evolutionary Computing
Neural solvers have demonstrated remarkable success in combinatorial optimization, often surpassing traditional heuristics in speed, solution quality, and generalization. However, their efficacy deteriorates significantly when confronted with complex constraints that cannot be effectively managed through simple masking mechanisms. To address this limitation, we introduce Universal Constrained Preference Optimization (UCPO), a novel plug-and-play framework that seamlessly integrates preference learning into existing neural solvers via a specially designed loss function, without requiring architectural modifications. UCPO embeds constraint satisfaction directly into a preference-based objective, eliminating the need for meticulous hyperparameter tuning. Leveraging a lightweight warm-start fine-tuning protocol, UCPO enables pre-trained models to consistently produce near-optimal, feasible solutions on challenging constraint-laden tasks, achieving exceptional performance with as little as 1\% of the original training budget.
title UCPO: A Universal Constrained Combinatorial Optimization Method via Preference Optimization
topic Neural and Evolutionary Computing
url https://arxiv.org/abs/2511.10148