Submitted by ChengyouJia 24 PaCo-RL: Advancing Reinforcement Learning for Consistent Image Generation with Pairwise Reward Modeling X-Gen Group 26 2