Multi-objective optimization models are often deployed in settings where decision makers must balance conflicting criteria. In practice, stakeholders must translate their qualitative preferences between the criteria, a step that can be difficult.
ParetoChat, described in a new arXiv paper, is an LLM-based interface designed for interactive multi-objective optimization. It aims to support this translation process, making it easier for stakeholders to work with such models.
The paper's abstract is brief, but it points to a real challenge in applying multi-objective optimization. By using a large language model as an interface, ParetoChat could help bridge the gap between human preferences and model objectives.