Yuhui Zhang

Hi! I am a postdoctoral scholar at Stanford University, working with Professors Serena Yeung-Levy, Ludwig Schmidt, and Emma Lundberg. Previously, I received my Ph.D. in Computer Science from Stanford University and my B.E. in Computer Science from Tsinghua University.

Research

My research focuses on the foundations and applications of multimodal intelligence. We live in a multimodal world, perceiving and reasoning through vision, language, sound, and other forms of information. This is especially important in science, including biology and medicine, where knowledge spans heterogeneous data modalities.

To build general physical and scientific intelligence, we need models that can reliably represent, understand, reason over, and generate across diverse modalities. My work pursues this goal along two complementary directions: developing foundational understandings of multimodal models, including modality gaps in CLIP [NeurIPS'22, ICLR'23, ICLR'24, Preprint'25], visual perception limits in multimodal language models [NeurIPS'24, CVPR'25, EMNLP'25, Preprint'26], and robust training of flow matching [CVPR'26, CVPR'26], and translating these insights into applications such as virtual cell modeling [ICML'25, Preprint'26, Preprint'26, Preprint'26] and scientific reasoning & discovery [NEJM AI'24, CVPR'24, CVPR'25, EACL'26, Preprint'26].

News

Awards

Services

Miscellaneous