Kazuki Ota

I am a Ph.D. student in Engineering at The University of Tokyo, advised by Takayuki Osa and Tatsuya Harada. I am also a part-time researcher at RIKEN AIP.

I am strongly interested in creating machines that bring novel insights to humanity. My goal is to build reinforcement-learning systems that discover strong strategies and valuable ideas from the rules of a domain alone.

LinkedIn / Google Scholar / GitHub / Email

ICML 2026

I will present two papers at ICML 2026: one at the main conference on self-play reinforcement learning, and one at the AI for Math Workshop on self-supervised theorem discovery. Please see my ICML 2026 page for details.

Research

I study reinforcement-learning systems that can search, test, and organize ideas beyond direct human supervision. My recent work explores this direction in two settings: stable and efficient self-play for discovering strong strategies in two-player games, and self-supervised theorem discovery from axioms alone in formal mathematical environments.

Reinforcement Learning Self-Play Search and Planning AI for Mathematics Theorem Discovery