ORLM is a novel framework that trains open-source large language models to automate complex optimization modeling tasks that traditionally require Operations Research experts.

• Reduces dependency on closed-source LLMs like GPT-4 for optimization tasks
• Creates specialized models that can generate optimization formulations and solver code
• Achieves competitive performance with customized training on OR-specific datasets
• Makes advanced optimization capabilities more accessible and affordable

This research democratizes Operations Research tools by enabling engineers to deploy cost-effective, specialized AI models for industrial optimization problems without extensive OR knowledge or reliance on proprietary models.

ORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling