Ku Onoda

Ku Onoda

M.S. student at the University of Tokyo, Matsuo-Iwasawa Laboratory.

Multi-Axis Max@K Reinforcement Learning for Representative Diversity in Text-to-Image Generation

Published in IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027), 2026

This work formalizes target-mode coverage and proposes multi-axis max@K, a group-based reinforcement learning objective for improving representative diversity in diffusion-based text-to-image models.

Paper · Code

Recommended citation: Ku Onoda, Paavo Parmas, Hiroki Furuta, Soichiro Nishimori, Yuta Oshima, Shohei Taniguchi, and Yutaka Matsuo.
Download Paper