← back to paper
arxiv: 2605.17602 · 2 revisions
AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image Alignment