About

I am an M.S. student in Technology Management for Innovation at the Graduate School of Engineering, The University of Tokyo, and a member of the Matsuo-Iwasawa Laboratory.

My research focuses on self-improving agent systems and reinforcement learning.

Selected research

• Multi-Axis Max@K Reinforcement Learning introduces a group-based reinforcement learning objective that improves target-mode coverage in text-to-image generation. Accepted at WACV 2027. [arxiv] [code]
• Does “Do Differentiable Simulators Give Better Policy Gradients?” Give Better Policy Gradients? studies gradient estimation in differentiable simulators and practical variance control. Published at ICLR 2026. [arxiv]
• AutoRef automatically optimizes the agentic harness around frozen image models for multi-reference image generation. [arxiv] [code]

For a complete overview, see Publications, Projects, or my CV.

Experience

• Third Intelligence — Research Scientist Intern, Feb 2026 - Present
• Matsuo Institute — Machine Learning Engineer Intern, Jan 2025 - Present

Beyond research

I have competed in rowing for ten years. I represented Japan at the 2023 World Rowing Under 23 Championships.

Links

Google Scholar · GitHub · X · LinkedIn