An Intuitive Explanation of Sparse Autoencoders for LLM Interpretability | Adam Karvonen

Sparse Autoencoders (SAEs) have recently become popular for interpretability of machine learning models (although sparse dictionary learning has been around since 1997). Machine learning models and LLMs are becoming more powerful and useful, but they are still black boxes, and we don’t understand how they do the things that they are capable of. It seems like it would be useful if we could understand how they work.

Sparse Autoencoders (SAEs) have recently become popular for interpretability of machine learning models (although sparse dictionary learning has been around since 1997 ). Machine learning models and LLMs are becoming more powerful and useful, but they are still black boxes, and we don’t understand how they do the things that they are capable of. It seems like it would be useful if we could understand how they work. Using SAEs, we can begin to break down a model’s computation into understandable components . There are several existing explanations of SAEs, and I wanted to create a brief writeup

saved by

• jeremy h
• Yoyo Yuan
• Lydia Nottingham
• Vincent Cheng
• Timothy Kostolansky
• Adi Mehta

related reading

• An Intuitive Explanation of Sparse Autoencoders for Mechanistic Interpretability of LLMs — LessWronglesswrong.com
• A gentle introduction to sparse autoencoders — LessWronglesswrong.com
• Towards Monosemanticity: Decomposing Language Models With Dictionary Learningtransformer-circuits.pub
• A gentle introduction to sparse autoencodersnickjiang.substack.com
• An Extremely Opinionated Annotated List of My Favourite Mechanistic Interpretability Papers v2 — AI Alignment Forumalignmentforum.org
• Sparse Autoencoders Find Highly Interpretable Features in Language Modelsarxiv.org
• pdfopenreview.net
• An Extremely Opinionated Annotated List of My Favourite Mechanistic Interpretability Papers v2 — AI Alignment Forumalignmentforum.org
• Interpretability with Sparse Autoencoders (Colab exercises) — LessWronglesswrong.com
• [2409.14507] A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencodersar5iv.labs.arxiv.org
• Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnettransformer-circuits.pub
• I Trained a Language Model. Then I Built a Brain Scanner and Looked Inside It. | by Caleb DeLeeuw | Mediummedium.com