LaT: LLM-as-Trainer for Multi-Task Vehicle Routing Solvers
This work addresses the training bias in multi-task VRP solvers by introducing a plug-and-play LLM-based training paradigm that enhances performance without bi-level optimization.
LaT uses a pretrained LLM as an external trainer to provide stage-wise guidance for multi-task vehicle routing solvers, improving solution quality across 16 VRP variants for several state-of-the-art models.
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.