Learning Native Reflection in Unified Models
Researchers present a method for unified multimodal models to jointly learn image generation and self-reflection text to repair their outputs.
Researchers present a method for unified multimodal models to jointly learn image generation and self-reflection text to repair their outputs.