๐ hi everyone,
i've recently been following a super cool research project coming from an anonymous twitter account xjdr. It's a novel approach to sampling ("the shrek sampler") that reproduces o-1 style reasoning using local 1B models. I made a video breaking it down for anyone that's curious
https://www.linkedin.com/posts/hchu1_the-most-exciting-llm-research-is-coming-activit[โฆ]706047696896-olyq?utm_source=share&utm_medium=member_desktop