About me

I am an Assistant Professor & PI at the National Institute of Informatics and SOKENDAI, and a Research Scientist at Sakana AI. I received my Ph.D. from UTokyo in 2025 and M.S. from PKU. Previously, I interned at Sakana AI, Adobe, and SenseTime.

My research centers on self-improving machine learning, with particular emphasis on generative models for vision and language. I pursue this through three directions:

• Modeling: I design efficient foundation models for vision and language tasks, including generative autoregressive and diffusion transformers, and novel architectures that better exploit structure in data and computation.
• Learning: I develop data-efficient learning algorithms that uncover and leverage hidden self-supervision in data, including semi-/self-/weakly supervised methods and robust training under limited or noisy labels.
• Adapting: I study neural networks in weight space, including model merging, composition, and generation. The goal is to understand and exploit the structure and knowledge encoded in model weights for instant adaptations.

Opportunity

I offer funded positions for interns, PhD students, and postdocs. If interested, email your CV, a brief research background, and the topics you would like to pursue in my group.

1. Interns through the NII International Internship Program with >100 univs.
2. Ph.D. students through the Informatics Program at SOKENDAI / NII.
3. Ph.D./Postdoc via DAAD IFI for students at German univs. and German nationals.
4. Research Assistants from universities in Japan at NII.

News

• 2025/11: I will be serving as an Area Chair at ICML 2026.
• 2025/04: I’ve joined NII and SOKENDAI as an Assistant Professor & Principal Investigator.
• 2025/03: Graduated from The University of Tokyo.
• 2025/01: I started an internship at Sakana AI as a Research Scientist, Tokyo.
• 2024/08: I gave an invited talk at MIRU 2024.