I'm an Assistant Professor of Computer Science at Princeton University. I received my Ph.D. from UC Berkeley, where I worked with Trevor Darrell, and my B.S. in Computer Science from Tsinghua University (Yao Class). Before joining Princeton, I was a research scientist at Meta FAIR in New York, and I have also spent time at Cornell, Intel, and Adobe.

I work on building models that see and understand the world, reason over long and multimodal context, and analyze complex inputs to solve challenging problems. My research spans deep learning and foundation models across vision, multimodal, and language. I value simplicity in the models, methods, and hypotheses that I explore.

In this spirit, some of my work has introduced simple, widely used designs in deep networks and efficient models (e.g., DenseNet, ConvNeXt, Network Slimming, Rethinking Pruning, Wanda). More recently, I have worked on multimodal and large language models, ranging from architectures (e.g., DyT, MetaMorph) to mechanism and data analysis (e.g., Massive Activations, dataset bias). My work has been recognized by 90,000 citations and a CVPR Best Paper Award.