Reproduce “OptiChat: Bridging Optimization Models and Practitioners with Large Language Models” 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
Hao Chen, Gonzalo Esteban Constante-Flores, Krishna Sri Ipsit Mantri, Sai Madhukiran Kompalli, Akshdeep Singh Ahluwalia, Can Li
OptiChat is a natural language dialogue system that bridges optimization models and practitioners, enabling intuitive interaction and interpretation using large language models. The system autonomously generates model descriptions and diagnoses infeasibility in under one minute and achieves an average accuracy of 88.9% in responding to complex queries by combining predefined functions with LLM-driven code generation.
Key insights
- A hybrid approach, combining predefined, domain-specific functions with LLM code generation, effectively mitigates hallucinations and improves accuracy for critical analytical tasks in optimization.
- Advanced LLMs like o3 demonstrate superior reasoning capabilities for open-ended code generation tasks (e.g., "why-not" queries), making them more suitable for dynamic and complex interactions.
- Structured preprocessing by an 'Illustrator' agent and targeted syntax reminders are crucial for LLMs to accurately interpret model components and significantly improve overall system performance and efficiency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.