Saining Xie
Saining Xie researches visual representations, model architectures and generative AI. His work includes a 2018 UC San Diego doctoral thesis, ConvNeXt and masked visual learning in 2022, and diffusion transformers with William Peebles. The DiT paper, first submitted in December 2022 and presented at ICCV 2023, studies transformer-based latent image generation. He also coauthored the 2024 Cambrian-1 multimodal model study. He received the 2025 PAMI Young Researcher Award.
- Now at
- New York University
- Works
- 97
- Citations
- 58,037
- h-index
- 38
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2018
Visual representation research
Xie wrote a UC San Diego doctoral thesis on deep representation learning with structural priors. His research concerns the design of visual representations and neural-network architectures.
-
2022
Revisiting convolutional networks
He coauthored A ConvNet for the 2020s, the ConvNeXt study listed at CVPR 2022. He also studied masked autoencoders and self-supervised visual learning. These projects address how visual models learn useful image features.
-
2022–2023
Diffusion with transformers
With William Peebles, Xie proposed diffusion transformers. Their December 2022 preprint replaces the usual U-Net backbone with a transformer operating on latent image patches. The paper studies how model size affects generation quality and was presented at ICCV 2023.
-
2024–2025
Multimodal models and recognition
Xie coauthored Cambrian-1, a vision-centered exploration of multimodal language models presented at NeurIPS 2024. In 2025, the Computer Vision Foundation named him for the PAMI Young Researcher Award.
Sources and credits
Photo: cs.nyu.edu; NYU Computer Science Department; creator not stated. License: Not stated by source. Source: cs.nyu.edu. Cropped to a square and resized. Ask us to remove or replace this photo.