Timaeus is merging into Resolution. Read the announcement

Structural Inference: Interpreting Small Language Models with Susceptibilities

We introduce "susceptibilities" along with a framework for applying these measurements to discovering structure inside models ("structural inference") and validate this in a small language model.
Authors
Garrett Baker=, George Wang=, Jesse Hoogland, Vinayak Pathak, Daniel Murfet
Timaeus · = Equal contribution
Published
April 25, 2025

Build on our work

Our tools for susceptibilities, local learning coefficients, and SGMCMC sampling are open source in the devinterp library.

Work with us

Timaeus has merged into Resolution. Open roles are now posted on the Resolution careers page.