SymbOmni: Evolving Agentic Omni Models via Symbolic Concept Learning

Jinxiu Liu, Jianru Li, Tanqing Kuang +4

Visual generation is increasingly ubiquitous in diverse domains, from text-to-image/video synthesis to multimodal interactive creation. Yet prevailing monolithic models remain fund…