The Synthetic component of MedInjection-FR was created to expand the dataset while maintaining controlled task diversity and data quality.It was generated exclusively for the training set using GPT-4o as the generation model and includes a balanced mix of multiple-choice (MCQ) and open-ended (OEQ) question–answer pairs derived from clinical and biomedical sources.