Yanxun Li

Achieving high visual fidelity while ensuring physical consistency in video generation could redefine the standards for simulating realistic interactions in AI-generated content.

Flow-based generative models disentangle concepts naturally during a pivotal "Instantiation Stage," offering a sweet spot for targeted manipulation.