Reproduce “Augmenting Operations Research with Auto-Formulation of Optimization Models from Problem Descriptions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Rindranirina Ramamonjison, Haley Li, Timothy T. Yu, Shiqi He, Vishnu Rengan, Amin Banitalebi-Dehkordi, Zirui Zhou, Yong Zhang
Researchers at Huawei Technologies Canada and the University of British Columbia developed an augmented intelligence system called OptGen that automatically generates mathematical optimization formulations from natural language descriptions. This system, which includes a novel LPWP dataset, demonstrated improved generalization for out-of-domain problems, achieving 0.63 accuracy and outperforming baseline models by 18%.
Key insights
- A two-stage mapping strategy with an intermediate representation (IR) improves the ability to learn the syntax and grammar of target formulations and generalizes better than direct text-to-table mapping.
- The copy mechanism is critical for accurately transferring variable names and data parameters from the input description, significantly reducing "hallucinations" and improving accuracy, particularly on out-of-domain data.
- Prompt-guided generation further enhances performance by focusing the model on relevant parts of the description, aiding in more precise declaration generation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.