Jefouree

The discoveries worth talking about each week.


← Back to Jefouree

Story permalink

arXiv Language Models

How diffusion models learn to un-scramble words while staying on target

Log in to share

It's like unscrambling a Scrabble board where halfway through, you pause to check: "Wait, am I building the right word?" and rearrange tiles that missed the goal.

This means we can now steer text-generation models toward specific rewards — like safety or factuality — without sacrificing creativity or speed.


Bug reported: No

Confirm action