Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization
Large language models (LLMs) are increasingly used to formulate optimization models from natural-language problem descriptions, yet realistic operations research (OR) requests are often incomplete: missing objectives, constraints, or business rules can change the resulting mathematical program. Existing evaluations largely assume a complete specification and therefore overlook whether an agent…
We haven't written up this one. arXiv cs.AI has the full story — the link below goes straight to it.
