TY  - RPRT
TI  - Post-Training Frontier Text-to-Image Models by Composing Preference and Rubric Rewards
AU  - Yuanhao Ban
AU  - I-Hung Hsu
AU  - Anastasios Angelopoulos
AU  - Wei-Lin Chiang
AU  - Ion Stoica
AU  - Cho-Jui Hsieh
PY  - 2026
UR  - https://arxiv.org/abs/2610.02967
ID  - 2610.02967
ER  -