TLDR: This post accompanies The Developmental Landscape of In-Context Learning by Jesse Hoogland, George Wang, Matthew Farrugia-Roberts, Liam Carroll, Susan Wei and Daniel Murfet (2024), which shows that in-context learning emerges in discrete, interpretable developmental stages, and that these stages can be discovered in a model- and data-agnostic way by probing the local geometry of the loss landscape.

Four months ago, we shared a discussion here of a paper which studied stagewise development in the toy model of superposition of Elhage et al. using ideas from Singular Learning Theory (SLT). The purpose of this document is to accompany a follow-up paper by Jesse Hoogland, George Wang, Matthew Farrugia-Roberts, Liam Carroll, Susan Wei and Daniel Murfet, which has taken a closer look at stagewise development in transformers at significantly larger scale, including language models, using an evolved version of these techniques.

How does in-context learning emerge? In [...]

---

Outline:

(06:21) On Stagewise Development

(08:30) Developmental Stages

(09:57) Developmental Milestones

(12:12) Discovering Stages

(12:33) The Local Learning Coefficient

(17:30) Essential Dynamics

(21:49) Implications

The original text contained 2 footnotes which were omitted from this narration.

---

First published:
March 20th, 2024

Source:
https://www.lesswrong.com/posts/Zza9MNA7YtHkzAtit/stagewise-development-in-neural-networks

---

Narrated by TYPE III AUDIO.

“Stagewise Development in Neural Networks” by Jesse Hoogland, Liam Carroll, Daniel Murfet

LessWrong (30+ Karma)