Healthy Cooking with Large Language Models, Supervised Fine-Tuning, and Retrieval Augmented Generation

In response to the growing demand of the global society to adopt healthy habits, this paper presents the design, development and validation of a Spanish culinary recipe dataset focused on healthy nutrition that includes representative dishes from the gastronomy of Spanish-speaking countries. We also evaluate the dataset using Gemma 2B, Supervised FineTuning and Retrieval Augmented Generation methodologies, showing its use to solve concrete problems related to nutrition.