The actual codebase of libreoffice is in the dataset, and a model is designed to replicate plausibility from its dataset