1 paper · 1 filter
Jinbin Bai, Yu Lei, Qingyu Shi +6
Aligning large visual generative models with human feedback is often performed through pairwise preference optimization. While such approaches are conceptually simple, they fundame…