TriVAL: A Tri-Validation Framework for Faithful Automatic Optimization Modeling

ArXi:2605.23966v1 Announce Type: cross Optimization modeling serves as the pivotal bridge between natural-language problem descriptions and optimization solvers, and remains a cornerstone for bringing operations research (OR) into real-world decision making. Recent advances in large language models (LLMs) have driven significant progress in automatic optimization modeling. However, existing methods still lack explicit validation during the modeling process, allowing errors