Skip to yearly menu bar Skip to main content


Poster

Lines of Thought in Large Language Models

Raphaël Sarfati · Toni Liu · Nicolas Boulle · Christopher Earls

Hall 3 + Hall 2B #323
[ ] [ Project Page ]
Wed 23 Apr 7 p.m. PDT — 9:30 p.m. PDT

Abstract:

Large Language Models achieve next-token prediction by transporting a vectorized piece of text (prompt) across an accompanying embedding space under the action of successive transformer layers. The resulting high-dimensional trajectories realize different contextualization, or 'thinking', steps, and fully determine the output probability distribution. We aim to characterize the statistical properties of ensembles of these 'lines of thought.' We observe that independent trajectories cluster along a low-dimensional, non-Euclidean manifold, and that their path can be well approximated by a stochastic equation with few parameters extracted from data. We find it remarkable that the vast complexity of such large models can be reduced to a much simpler form, and we reflect on implications.

Live content is unavailable. Log in and register to view live content