The Inside View cover image

Collin Burns On Discovering Latent Knowledge In Language Models Without Supervision

The Inside View

CHAPTER

The Key Findings of Zero Shot Prompting

The method of taking in unlabeled hidden states from language model and trying to classify them as rarefals this actually just gets high accuracy. It even slightly outperforms zero shot prompting so that's when you basically take a prompt and ask a model like okay consider the following review is it positive or negative? And then look at the probability of the next token and see is the next token positive or negative  and use that as the prediction.

00:00
Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner