Bridging Synthetic and Real Routing Problems via LLM-Guided Instance Generation and Progressive Adaptation

2025-11-13

AI Summary

Neural combinatorial optimization (NCO) models achieve strong performance on synthetic data but suffer from poor generalization to real-world vehicle routing problems (VRPs), such as TSPLib and CVRPLib benchmarks. To address this, we propose EvoReal—a novel framework that pioneers the integration of large language models (LLMs) into evolutionary instance generation, enabling the synthesis of training data with realistic structural characteristics. EvoReal further employs statistical distribution alignment and a phased fine-tuning strategy to realize progressive domain adaptation from synthetic to real VRP instances. Our approach significantly narrows the performance gap of NCO models on real-world VRPs: average optimality gaps reduce to 1.05% on TSPLib and 2.71% on CVRPLib, relative to optimal or best-known solutions. These results empirically validate the effectiveness of LLM-guided synthetic data generation coupled with progressive domain adaptation in enhancing the generalization capability of NCO models.

Technology Category

• Search and Optimization: Evolutionary Computation
• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Evolutionary Learning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

Recent advances in Neural Combinatorial Optimization (NCO) methods have significantly improved the capability of neural solvers to handle synthetic routing instances. Nonetheless, existing neural solvers typically struggle to generalize effectively from synthetic, uniformly-distributed training data to real-world VRP scenarios, including widely recognized benchmark instances from TSPLib and CVRPLib. To bridge this generalization gap, we present Evolutionary Realistic Instance Synthesis (EvoReal), which leverages an evolutionary module guided by large language models (LLMs) to generate synthetic instances characterized by diverse and realistic structural patterns. Specifically, the evolutionary module produces synthetic instances whose structural attributes statistically mimics those observed in authentic real-world instances. Subsequently, pre-trained NCO models are progressively refined, firstly aligning them with these structurally enriched synthetic distributions and then further adapting them through direct fine-tuning on actual benchmark instances. Extensive experimental evaluations demonstrate that EvoReal markedly improves the generalization capabilities of state-of-the-art neural solvers, yielding a notable reduced performance gap compared to the optimal solutions on the TSPLib (1.05%) and CVRPLib (2.71%) benchmarks across a broad spectrum of problem scales.

Problem

• Bridging generalization gap between synthetic and real routing problems
• Generating realistic synthetic instances using LLM-guided evolution
• Progressively adapting neural solvers to real-world benchmark instances

Innovation

• LLM-guided evolutionary module generates realistic synthetic instances
• Progressively refines pre-trained NCO models with synthetic distributions
• Further adapts models through fine-tuning on actual benchmark instances