Skip to content

Anthropic company blog

Decomposing language models into understandable components

by @AnthropicAI anthropic.com published 2023-10-05

Interpretability research decomposing a transformer layer into thousands of interpretable features that can steer model behavior.

Read on anthropic.com → Discuss on HN (62) Archived copy

Points: 445
HN comments: 62
Submissions: 1
Cited by: 0 commenters

AI & Machine Learning Math & Algorithms 2023

On Hacker News

Submitted	Title on HN	Points	Comments
Oct 8, 2023	Decomposing language models into understandable components by tompark	445	62

More from Anthropic

All 146 posts →

Statement on US government directive to suspend access to Fable 5 and Mythos 5

20263,158 points2,314 commentscited by 11AI & Machine Learning details

Statement from Dario Amodei on our discussions with the Department of War

20262,920 points1,579 commentscited by 14Society, Policy & Culture details

Claude Fable 5

20262,626 points2,159 commentscited by 10AI & Machine Learning details

Claude Opus 4.6

20262,346 points1,031 commentscited by 9AI & Machine Learning details

Claude 3.7 Sonnet and Claude Code

20252,127 points963 commentscited by 6AI & Machine Learning details

Claude 4

20252,013 points1,170 comments AI & Machine Learning details

Related posts from the same era

Bing: “I will not harm you unless you harm me first”

Simon Willison20233,363 points1,673 commentscited by 6 details

GPT-4o

OpenAI Blog20243,138 points2,366 commentscited by 16 details

Open source AI is the path forward

Meta Newsroom20242,360 points887 commentscited by 5 details

Building a Virtual Machine Inside ChatGPT

Jonas Degrave20222,029 points908 commentscited by 16 details

The man who killed Google Search?

Ed Zitron20241,884 points878 commentscited by 47 details

ChatGPT Plugins

OpenAI Blog20231,875 points1,106 commentscited by 14 details