Sheng Li

Hi! I am Sheng Li. I am currently a final-year Ph.D. student in Department of Computer Science at the University of Pittsburgh, advised by Dr. Xulong Tang. I received my B.E. degree in Software Engineering from Sichuan University, China, in 2020.

My research focuses on designing and optimizing Efficient Machine Learning Algorithms and Systems, spanning applications in Diffusion-based Video Generation, RAG-enabled LLM Serving, Self-supervised Contrastive Learning, Supervised Model Training, and 3D Gaussian Splatting. My research targets platforms ranging from large-scale cloud clusters to resource-constrained edge devices.

News

03/2026	One first-author paper on Efficient On-device Continual Learning is submitted to TC!
03/2026	One first-author paper on Efficient Video Diffusion is submitted to ECCV2026!
02/2026	One paper is accepted to CVPR 2026. This is the intern project in Adobe! Thanks to all collaborators!
05/2025	Joining Adobe Research as a Research Scientist/Engineer Intern!
01/2025	One paper is accepted to ICLR 2025. Thanks to all collaborators!
09/2024	One paper is accepted to ASP-DAC 2025. Thanks to all collaborators!

Experiences

Adobe, May 2025 – November 2025
Research Scientist/Engineer Intern
Mentor: Dr. Yifan Gong
Topic: Efficient Diffusion-based Video Generation

Selected Publications

CVPR

Content-Aware Dynamic Patchification for Efficient Video Diffusion

Sheng Li, Connelly Barnes, Mamshad Nayeem Rizve, Hongwu Peng, Zhengang Li, Ohi Dibua, Alireza Ganjdanesh, Xulong Tang, Yan Kang, and Yifan Gong

In The IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026

ICLR

Mutual Effort for Efficiency: A Similarity-based Token Pruning for Vision Transformers in Self-Supervised Learning

Sheng Li^*, Qitao Tan^*, Yue Dai, Zhenglun Kong, Tianyu Wang, Jun Liu, Ao Li, Ninghao Liu, Yufei Ding, Xulong Tang, and Geng Yuan

In The Thirteenth International Conference on Learning Representations, 2025

Bib HTML

@inproceedings{li2025mutual,
  title = {Mutual Effort for Efficiency: A Similarity-based Token Pruning for Vision Transformers in Self-Supervised Learning},
  author = {Li, Sheng and Tan, Qitao and Dai, Yue and Kong, Zhenglun and Wang, Tianyu and Liu, Jun and Li, Ao and Liu, Ninghao and Ding, Yufei and Tang, Xulong and Yuan, Geng},
  booktitle = {The Thirteenth International Conference on Learning Representations},
  year = {2025},
  url = {https://openreview.net/forum?id=GTcEe5fayC}
}

ICLR

Waxing-and-Waning: a Generic Similarity-based Framework for Efficient Self-Supervised Learning

Sheng Li, Chao Wu, Ao Li, Yanzhi Wang, Xulong Tang, and Geng Yuan

In The Twelfth International Conference on Learning Representations, 2024

Bib HTML

@inproceedings{li2024waxingandwaning,
  title = {Waxing-and-Waning: a Generic Similarity-based Framework for Efficient Self-Supervised Learning},
  author = {Li, Sheng and Wu, Chao and Li, Ao and Wang, Yanzhi and Tang, Xulong and Yuan, Geng},
  booktitle = {The Twelfth International Conference on Learning Representations},
  year = {2024},
  url = {https://openreview.net/forum?id=TilcG5C8bN}
}

ICLR

SmartFRZ: An Efficient Training Framework using Attention-Based Layer Freezing

Sheng Li^*, Geng Yuan^*, Yue Dai^*, Youtao Zhang, Yanzhi Wang, and Xulong Tang

In The Eleventh International Conference on Learning Representations, 2023

Spotlight Bib HTML

@inproceedings{li2023smartfrz,
  title = {Smart{FRZ}: An Efficient Training Framework using Attention-Based Layer Freezing},
  author = {Li, Sheng and Yuan, Geng and Dai, Yue and Zhang, Youtao and Wang, Yanzhi and Tang, Xulong},
  booktitle = {The Eleventh International Conference on Learning Representations},
  year = {2023},
  url = {https://openreview.net/forum?id=i9UlAr1T_xl}
}

NeurIPS

Layer freezing & data sieving: missing pieces of a generic framework for sparse training

Geng Yuan^*, Yanyu Li^*, Sheng Li, Zhenglun Kong, Sergey Tulyakov, Xulong Tang, Yanzhi Wang, and Jian Ren

Advances in Neural Information Processing Systems, 2022

Bib HTML

@article{yuan2022layer,
  title = {Layer freezing \&amp; data sieving: missing pieces of a generic framework for sparse training},
  author = {Yuan, Geng and Li, Yanyu and Li, Sheng and Kong, Zhenglun and Tulyakov, Sergey and Tang, Xulong and Wang, Yanzhi and Ren, Jian},
  journal = {Advances in Neural Information Processing Systems},
  volume = {35},
  pages = {19061--19074},
  year = {2022},
  url = {https://proceedings.neurips.cc/paper_files/paper/2022/file/794a425a2e47e05d29d30f79b79a692d-Paper-Conference.pdf}
}