New preprint: TUNA — Taming Unified Visual Representations for Native Unified Multimodal Models (arXiv:2512.02014).