Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
LMs Can Teach Themselves to Think Before Speaking This paper presents a generalization of STaR, called Quiet-STaR, to enable language models (LMs) to learn to reason in more general and scalable ways. Quiet-STaR enables LMs to generate rationales at each token to explain… https://t.co/DrGfztxmXw
Quiet-STaR teaches language models to reason from arbitrary text by generating explanatory thoughts before predicting, enabling zero-shot reasoning abilities without curated data. A step towards bridging the reasoning gap between LMs and humans. https://t.co/NHKCLrE66m
Language models today are trained to reason either 1) generally, imitating online reasoning data or 2) narrowly, self-teaching on their own solutions to specific tasks Can LMs teach themselves to reason generally?🌟Introducing Quiet-STaR, self-teaching via internal monologue!🧵 https://t.co/WCSxLPZeCX
A recent development in artificial intelligence, Quiet-STaR, represents a significant advancement in the capabilities of language models (LMs). This new approach, developed in 2024, allows LMs to 'think before speaking,' mirroring the human process of pausing to think during conversation or writing. Quiet-STaR, a generalization of the STaR model, enables LMs to reason in a more general and scalable manner by generating rationales at each token to explain their reasoning process. The innovation, developed by researchers E Zelikman, G Harik, Y Shao, V Jayasiri, N Haber, and N D. Goodman from Stanford University & Notbad AI Inc, aims to bridge the reasoning gap between LMs and humans by teaching LMs to generate explanatory thoughts before predicting, thus enabling zero-shot reasoning abilities without the need for curated data.
