Large language models improve vehicle routing error detection with self generated code
Self-Guided LLM Reasoning for Error Diagnosis in Vehicle Routing Problems
Artificial IntelligenceMachine Learning
Summary
Vehicle routing problems involve planning the best routes for vehicles to deliver goods efficiently, but the data used often has errors. The authors studied how well large language models (LLMs) can find and fix these errors. They found that LLMs struggle with complex or tricky mistakes. To fix this, they designed a system where the LLM writes its own code to help analyze the problems better. This approach greatly improved the model's ability to spot errors almost perfectly.
What this means in practice
- •For logistics planners: Detect errors in vehicle routing data automatically to prevent costly routing mistakes in delivery operations.
- •For fleet management teams: Improve reliability of routing software by integrating LLM-based error diagnosis that generates debugging support code.
Authors
Ni Zhang, Jianan Zhou, Yue-Jiao Gong, Zhiguang Cao
Abstract
Vehicle Routing Problems (VRPs) are fundamental to many real-world applications. Recent advances in large language models (LLMs) have spurred progress in solving these NP-hard problems. However, most existing approaches assume that input instances are error-free, neglecting the presence of faults commonly encountered in real-world data. In this work, we systematically investigate the capability of LLMs to detect (and correct) errors in VRP instances, and identify key limitations, particularly in handling complex instances or errors that substantially increase the LLM's reasoning burden. To overcome these challenges, we propose a self-guided reasoning framework in which LLMs autonomously generate code that provides structured insights to aid in error diagnosis. Experiments across diverse problem settings and LLMs demonstrate that our framework significantly improves error detection performance, achieving near-perfect accuracy.