Post-Training Frontier Text-to-Image Models by Composing Preference and Rubric Rewards Paper • 2610.02967 • Published 8 days ago • 24