UniRect-CoT: Enhancing Generation in Unified Multimodal Models via Reflective Rectification with Inherent Understanding
Unified Rectification-CoT is proposed, a training-free unified rectification chain-of-thought framework that generalizes across existing flow-based UMMs, consistently enhancing overall text-to-image generation performance.
Yi-Bo Jiang, Tao Wu, Rui Jiang et al.
· 0 citations