Blog

half-baked contemplations

A Gentle Introduction to Mechanistic Interpretability

what a cell is to biological life, a "feature" is to a neural network

Finding Multidimensional Language Model Representations Using Persistent Homology

persistence is key to discovering interesting features

Singular Learning Theory: Basics

informal notes on the basics of singular learning theory

Empirical SLT: LLC Estimation, Susceptibilities, and Structural Inference

informal notes: estimating the local learning coefficient, susceptibilities and spectroscopy