Gurnee & Tegmark (2023) trained linear probes to take an LLM's internal activation on a landmark's name (e.g. "The London Eye"), and predict the landmark's longitude and latitude. The results look like this:[1]

Two angles of true world atlas, with predicted atlas hovering above. True locations are red points; predicted locations are blue, in a slightly raised plane, linked to the corresponding true location by a grey line.

So LLMs (or at least, Llama 2, which they used for this experiment) contain a pretty good linear representation of an atlas.

Sometimes, like when thinking about distances, a globe is more useful than an atlas. Do models use the globe representation? To find out, we can train probes to predict the (x,y,z) coordinates of landmarks, viewed as living in 3D space. Here are the results:

Left: Europe faces us. Middle: The Pacific faces us. Right: India faces us. Predicted points [...]

The original text contained 1 footnote which was omitted from this narration.

The original text contained 3 images which were described by AI.

---

First published:
October 8th, 2024

Source:
https://www.lesswrong.com/posts/guNzr32FC6DYCaSgJ/there-is-a-globe-in-your-llm

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

“There is a globe in your LLM” by jacob_drori

LessWrong (30+ Karma)