"Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization" by Sihan Ge , Yichen Lin , Chenyu Zhou , Jianghao Lin , Tao Yao , Dongdong Ge
TLDR:
The text discusses the challenges faced when using large language models (LLMs) to create optimization models from incomplete natural-language problem descriptions in operations research (OR). Existing evaluations often overlook the need for clarification before modeling begins. To address this issue, the text introduces OR-Clarify, a benchmark for pre-formulation clarification that assesses agents' ability to extract missing information from partial problem descriptions. Additionally, a two-stage framework called InterOPT is proposed to guide agents on when to ask questions or stop based on missing critical details in the formulation. Results show that in choice-based experiments, InterOPT outperforms baselines in recovering exact information, highlighting the importance of selective completeness in OR assistance.
Read Paper / Blog
huggingface.co
Paper page - Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization
Join the discussion on this paper page
1September 7, 2026 38
