FoodPuzzle: Developing Large Language Model Agents as Flavor Scientists

TL;DR

This paper introduces FoodPuzzle, a new benchmark and scientific agent approach using large language models to improve flavor profile prediction and hypothesis generation in food science, aiming to accelerate and scale flavor development.

Contribution

The paper presents a novel scientific agent framework with in-context learning and retrieval techniques, along with a large benchmark dataset for flavor science, advancing AI-driven flavor development.

Findings

Model significantly outperforms traditional flavor prediction methods.

The FoodPuzzle benchmark enables rigorous evaluation of flavor hypothesis generation.

Proposed approach demonstrates potential to transform flavor development practices.

Abstract

Flavor development in the food industry is increasingly challenged by the need for rapid innovation and precise flavor profile creation. Traditional flavor research methods typically rely on iterative, subjective testing, which lacks the efficiency and scalability required for modern demands. This paper presents three contributions to address the challenges. Firstly, we define a new problem domain for scientific agents in flavor science, conceptualized as the generation of hypotheses for flavor profile sourcing and understanding. To facilitate research in this area, we introduce the FoodPuzzle, a challenging benchmark consisting of 978 food items and 1,766 flavor molecules profiles. We propose a novel Scientific Agent approach, integrating in-context learning and retrieval augmented techniques to generate grounded hypotheses in the domain of food science. Experimental results indicate that our model significantly surpasses traditional methods in flavor profile prediction tasks, demonstrating its potential to transform flavor development practices.