Baining Guo
Also known as 郭百宁
Baining Guo is a computer vision researcher whose work includes Swin Transformer and AI for robotic manipulation. He earned degrees at Beijing University and Cornell, joined Microsoft in 1999 after working at Intel, and chaired SIGGRAPH Asia's technical papers program in 2014. Swin Transformer won ICCV's 2021 Marr Prize. In November 2024, he coauthored CogACT for robot manipulation; in 2025, he coauthored an FP4 training framework for large language models; in August 2026, he coauthored a paper on adaptive action horizons. As of October 2026, he is a Distinguished Scientist at Microsoft Research and a Technical Fellow and Deputy Managing Director at Microsoft Research Asia.
- Now at
- Microsoft Research
Timeline
-
1982–1991
Degrees from Beijing and Cornell
He earns a B.S. from Beijing University in 1982, then earns M.S. and Ph.D. degrees from Cornell University in 1989 and 1991.
-
1999
Joining Microsoft from Intel
He joins Microsoft in 1999 after working as a senior staff researcher with Intel Corporation's Microcomputer Research Labs in Santa Clara, California.
-
2014
Technical Papers Chair at SIGGRAPH Asia
At SIGGRAPH Asia 2014, he serves as Technical Papers Chair for Microsoft Research Asia.
-
2021
Swin Transformer wins the Marr Prize
He coauthors Swin Transformer in 2021, a hierarchical vision Transformer that uses shifted windows to limit self-attention to local windows. The model serves as a general-purpose backbone for computer vision, and the paper wins ICCV's Marr Prize.
-
November 2024
CogACT for robotic manipulation
He coauthors CogACT in November 2024, a vision-language-action model for robotic manipulation. Its architecture includes a specialized action module conditioned on vision-language model output.
-
2025
FP4 training for language models
He coauthors Optimizing Large Language Model Training Using FP4 Quantization in 2025. The paper introduces an FP4 training framework for large language models.
-
August 2026
Continue or Replan?
In August 2026, he coauthors Continue or Replan?, which proposes a Bernoulli-Continuation Policy for adaptive-horizon execution in vision-language-action models.
-
As of October 2026
Technical Fellow at Microsoft Research Asia
He is a Technical Fellow and Deputy Managing Director at Microsoft Research Asia. He is also a Distinguished Scientist at Microsoft Research.
Career
- Since 1999Microsoft ResearchDistinguished Scientist
- Until 1999IntelSenior staff researcher
Education
- 1991Cornell UniversityPh.D.
- 1989Cornell UniversityM.S.
- 1982Peking UniversityB.S.
Sources and credits
Photo: We have no published photo of Baining Guo yet. Send us a free-licensed photo.
- Baining Guo
- Asia Faculty Summit 2012 - Microsoft Research
- SIGGRAPH Asia 2014 Technical Papers
- Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
- Baining Guo at Microsoft Research - Publications
- CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation
- Optimizing Large Language Model Training Using FP4 Quantization
- Continue or Replan? Bernoulli-Continuation Policy Learning for Adaptive Horizon Execution
- Baining Guo at Microsoft Research
- Microsoft Research Asia - Microsoft Research: About Us



