Yikun Bai
Physical AI Postdoctoral Fellow @ Purdue University
Let all things be done decently and in order.
— 1 Corinthians 14:40
About Me
I am currently a Physical AI Postdoc Fellow at Purdue University, supervised by Dr. Guang Lin and Dr. Ruqi Zhang. Prior to Purdue, I was a postdoc in the Department of Computer Science at Vanderbilt University (2022-2025), supervised by Dr. Soheil Kolouri, my academic advisor. I received my Ph.D. in Electrical and Computer Engineering from the University of Delaware in 2022, where I studied machine learning and statistics under the guidance of Dr. Dominique Guillot.
Current Research Interests
- Generative Models: Flow matching and diffusion models, with emphasis on probability-path design, transport dynamics, and scalable sampling for high-dimensional and non-Euclidean data.
- Reinforcement Learning: RL post-training, reasoning, and generative-model fine-tuning for controllable and reliable generation.
- Geometric / Mathematical Machine Learning: Geometry-aware learning and non-Euclidean generative models, including optimal transport and structure-aware modeling.
News
| Mar 15, 2026 | Our paper Generalized Discrete Diffusion with Self-Correction has been accepted by ICML 2026. |
|---|---|
| Jun 1, 2024 | One paper, Stereographic Spherical Sliced Wasserstein Distances, was accepted by ICML 2024. |
| Feb 15, 2024 | Four papers were accepted by ICLR 2024: |
| Nov 1, 2023 | Our paper LCOT: Linear Circular Optimal Transport was accepted by ICLR 2023. |
| Oct 24, 2023 | Presentation at Korea Institute for Advanced Study — More Info. |