Abstract page for arXiv paper 2508.02317: VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo