Mean field theory

with Lauren Greenspan · in progress

An introduction to mean field theory for neural networks: the mean-field (critical feature-learning) scaling, how it differs from the lazy/NTK regime, and what it predicts about learning and generalization. The series is published on LessWrong; this page collects the posts and related resources hosted here.

The series

  1. An introduction LessWrong · April 2026
  2. Part 2 — in preparation

Resources