TY  - RPRT
TI  - Rethinking the Design Space of Reinforcement Learning for Diffusion Models: On the Importance of Likelihood Estimation Beyond Loss Design
AU  - Jaemoo Choi
AU  - Yuchen Zhu
AU  - Wei Guo
AU  - Petr Molodyk
AU  - Bo Yuan
AU  - Jinbin Bai
AU  - Yi Xin
AU  - Molei Tao
AU  - Yongxin Chen
PY  - 2026
UR  - https://arxiv.org/abs/2602.04663
ID  - 2602.04663
ER  -