InterpretabilityResearch

Toy Models of Superposition

Sep 14, 2022
Read Paper

Abstract

In this paper, we use toy models — small ReLU networks trained on synthetic data with sparse input features — to investigate how and when models represent more features than they have dimensions. We call this phenomenon superposition. When features are sparse, superposition allows compression beyond what a linear model would do, at the cost of "interference" that requires nonlinear filtering.

Related content

Introducing Bloom: an open source tool for automated behavioral evaluations

Read more

Project Vend: Phase two

In June, we revealed that we’d set up a small shop in our San Francisco office lunchroom, run by an AI shopkeeper. It was part of Project Vend, a free-form experiment exploring how well AIs could do on complex, real-world tasks. How has Claude's business been since we last wrote?

Read more

Introducing Anthropic Interviewer: What 1,250 professionals told us about working with AI

We built an interview tool called Anthropic Interviewer. Powered by Claude, Anthropic Interviewer runs detailed interviews automatically and at unprecedented scale.

Read more
Toy Models of Superposition \ Anthropic