Diffusion Large Language Models (dLLMs) can write many parts of an answer at once, not just left to right like usual chatbots.
This paper teaches AI to notice not just what is in a picture, but how the picture looks and feels to people.