zroe1/toy-models-of-superposition

A replication of "Toy Models of Superposition," a groundbreaking machine learning research paper published by authors affiliated with Anthropic and Harvard in 2022.

13
/ 100
Experimental

This project investigates how small AI models can represent more features than they seemingly have capacity for, a concept known as "superposition." It takes the experimental setup and findings from a prominent research paper and recreates them, providing a detailed PDF report and executable code. AI researchers, particularly those studying model interpretability or efficiency, would use this to understand and reproduce these specific findings.

No commits in the last 6 months.

Use this if you are an AI researcher or student who wants to understand, replicate, and experiment with the 'superposition' phenomenon in toy machine learning models described by Anthropic and Harvard.

Not ideal if you are looking for a practical tool to apply to real-world datasets or for a broad, conceptual overview of AI interpretability.

AI-interpretability neural-network-mechanisms machine-learning-research model-capacity sparse-coding
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 5 / 25
Maturity 8 / 25
Community 0 / 25

How are scores calculated?

Stars

9

Forks

Language

Jupyter Notebook

License

Last pushed

Dec 30, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/zroe1/toy-models-of-superposition"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.