We are happy that we just published a new preprint on arxiv about our latest improvement in diffusion models for synthetisizing IMU data. In this work, we were able to significantly reduce the amount of training epochs without a performance reduction in the classification task. This saves ressources, and reduces time to train diffusion models.