DiT-Reward: Generative Representations for Text-to-Image Reward Modeling

arXiv·medium signal

DiT-Reward asks whether representations learned for image generation can also evaluate generated images, repurposing generative features for reward modeling in text-to-image systems. For teams doing RLHF or preference tuning on image models, reusing the generator's own representations could cut the cost of building a separate reward stack.