Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2

Lieberum, Rajamanoharan, Conmy et al. (DeepMind) (2024)

Read paper

Tags: open-source, tooling, deepmind, gemma

Abstract

We release Gemma Scope, a comprehensive suite of open sparse autoencoders trained on every layer and sublayer of Gemma 2 models, providing the community with tools for interpretability research.