DAOpt: Modeling and Evaluation of Data-Driven Optimization under Uncertainty with LLMs
This work addresses the gap in using LLMs for uncertain decision-making in optimization, which is incremental as it builds on existing deterministic methods.
The paper tackles the problem of applying large language models (LLMs) to uncertain optimization settings, which are underexplored, by proposing the DAOpt framework with a dataset, multi-agent module, and simulation environment, resulting in improved modeling capabilities through few-shot learning.
Recent advances in large language models (LLMs) have accelerated research on automated optimization modeling. While real-world decision-making is inherently uncertain, most existing work has focused on deterministic optimization with known parameters, leaving the application of LLMs in uncertain settings largely unexplored. To that end, we propose the DAOpt framework including a new dataset OptU, a multi-agent decision-making module, and a simulation environment for evaluating LLMs with a focus on out-of-sample feasibility and robustness. Additionally, we enhance LLMs' modeling capabilities by incorporating few-shot learning with domain knowledge from stochastic and robust optimization.