arXiv cs.AIFriday · May 29, 2026FREE

Opt-Verifier: Unleashing the Power of LLMs for Optimization Modeling via Dual-Side Verification

Published on arXiv cs.AI on May 29, 2026, Opt-Verifier is a novel framework that leverages large language models (LLMs) to automate the creation of mathematical optimization models while significantly improving their verification. Traditional methods for building these models in operations research demand substantial human expertise, and while LLMs have begun to automate this, they often struggle to verify the correctness of their outputs. Existing LLM approaches frequently fail to check the rationality of constraints, variables, or the validity of solutions, which hinders subsequent correction and impacts modeling accuracy. Opt-Verifier addresses this by implementing a "Dual-side Verification" mechanism. The first component, structure-side verification, ensures that the generated optimization model's structure accurately aligns with the original problem description, capturing all constraints and requirements precisely. The second component, solution-side verification, interprets and evaluates the validity of the model's solutions, confirming their logical and mathematical soundness. By integrating these two verification perspectives, Opt-Verifier aims to overcome the limitations of current LLM-based modeling tools, thereby enhancing the overall accuracy and reliability of automatically generated optimization models. This research aims to make AI-driven optimization modeling more robust and trustworthy.