LaT: LLM-as-Trainer for Multi-Task Vehicle Routing Solvers
2026-07-20 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors address the challenge of training a single neural model to solve many types of Vehicle Routing Problems (VRPs) that have different rules and difficulty levels. They propose a new training method called LLM-as-Trainer (LaT), which uses a large language model to give the neural solver guidance based on how well it is doing across tasks. This extra help improves the model's performance on both the VRP types it was trained on and new ones it hasn't seen before. Their experiments show LaT works well without needing costly extra training steps.
Vehicle Routing Problemmulti-task learningneural solverlarge language modelmeta-learningpolicy optimizationtraining paradigmconstraint vectorcross-task validation
Authors
Yang Wang, Ya-Hui Jia, Wei-Neng Chen, Yi Mei, Wen Song, Zhiguang Cao
Abstract
Multi-task neural solvers aim to handle multiple Vehicle Routing Problem (VRP) variants within a unified model, avoiding separate training for each constraint combination. However, VRP variants differ in optimization difficulty, while existing methods lack stage-wise feedback on their training status, making the model biased to some specific variants. Although meta-learning can support adaptive training, it typically requires bi-level optimization and additional gradient updates, increasing computational cost. To address this limitation, we propose LLM-as-Trainer (LaT), a plug-and-play training paradigm that uses a pretrained large language model as an external trainer. LaT periodically analyzes cross-task validation metrics to generate a stage-wise guidance vector. This vector is combined with the current task's constraint vector and injected into each encoder layer, providing the neural solver with additional training information during subsequent policy optimization. Experiments on 16 VRP variants show that LaT improves the solution quality of several state-of-the-art multi-task neural solvers on both trained and unseen variants, supporting the effectiveness and generality of the proposed training paradigm.