The paper in question: https://arxiv.org/pdf/2310.16834

https://www.youtube.com/watch?v=K_9wQ6LZNpI

Some modern LLMs based on diffusion:

• Dream 7B: https://arxiv.org/pdf/2508.15487?
• Mercury 2 reasoning model: https://www.inceptionlabs.ai/blog/introducing-mercury-2