"Large Flow Language Models"

What if we could turn existing LLMs into diffusionLMs that generate multiple tokens in parallel without losing accuracy?

📓 Read

@datascienceiot