Generative Urban Foundation Model

Decoder-only transformer generating realistic human-mobility trajectories across world cities from real data

Investigator on the Generative Urban Foundation Model in Sekimoto Lab (IIS, CSIS), The University of Tokyo — generating realistic human-mobility trajectories across global cities from real data with a decoder-only transformer, and exploring world models and diffusion methods.

This is current research (2025–present). The academic write-up lives on the research one-pager.

Focus

  • Foundation models for human mobility
  • Decoder-only transformers
  • World models and diffusion methods