@adamrpearce
github.com/1wheel
[ my public key: https://keybase.io/1wheel; my proof: https://keybase.io/1wheel/sigs/oiIWaeFlocOXCQTRs_2pfsieq2ldvt0hIwB9xoUwtbc ]
user: 1wheel
1928 karma | created 14 years ago
recent submissions
When Models Manipulate Manifolds: The Geometry of a Counting Task
(transformer-circuits.pub)
4 pts|4 months ago|discuss
9 months ago|discuss
10 months ago|discuss
1 year ago|discuss
Advent of Code Solve Times
(roadtolarissa.com)
2 pts|1 year ago|discuss
Anthropic raises an additional $4B from Amazon
(techcrunch.com)
4 pts|1 year ago|discuss
1 year ago|discuss
Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet
(transformer-circuits.pub)
168 pts|1 year ago|124 comments
The Claude 3 Model Family: Opus, Sonnet, Haiku [pdf]
(www-cdn.anthropic.com)
33 pts|2 years ago|3 comments
Circuits Updates – January 2024
(transformer-circuits.pub)
1 pts|2 years ago|discuss
Patchscopes: A Unifying Framework for Inspecting Hidden Representations of LMs
(pair-code.github.io)
2 pts|2 years ago|discuss
2 years ago|discuss
2 years ago|discuss
Do Machine Learning Models Memorize or Generalize?
(pair.withgoogle.com)
454 pts|2 years ago|210 comments
An interactive introduction to grokking and mechanistic interpretability
(pair.withgoogle.com)
1 pts|2 years ago|discuss
From Confidently Incorrect Models to Humble Ensembles
(pair.withgoogle.com)
1 pts|2 years ago|1 comment
2 years ago|discuss
Speeding up the JS ecosystem - npm scripts
(marvinh.dev)
9 pts|3 years ago|3 comments
3 years ago|discuss
Searching for Unintended Biases with Saliency
(pair.withgoogle.com)
2 pts|3 years ago|1 comment