Renaissance Labs

Atlas / North America

Nonprofit · San Francisco, United States

Neuronpedia

An open-source interpretability platform for exploring, visualizing and steering model internals such as sparse autoencoder features.

interpretabilitysparse autoencoderstooling

San Francisco, United States

Founded
2023
Type
Nonprofit
Based in
San Francisco, United States
Featured projects
4

About

Neuronpedia was created by Johnny Lin, an engineer who previously worked at Apple. It is an open-source platform where researchers and the public can search the features inside language models, steer model behavior and trace circuits.

It hosts feature dashboards and tooling for releases such as Gemma Scope 2, Llama Scope and SAE Bench, and recent additions include Circuit Tracer, the Assistant Axis, Natural Language Autoencoders and Jacobian Lens. It also offers an API and a community Slack.

Funding comes from Decode Research, Open Philanthropy, the Long Term Future Fund and others, and its users include groups at Anthropic, Google DeepMind, EleutherAI and Apollo Research.

Mission and ambitions

Neuronpedia aims to make model internals easy to inspect and share, so that interpretability research is open and collaborative.

Known for

  1. 01

    Public sparse autoencoder feature explorer

  2. 02

    Hosting Gemma Scope and Llama Scope

  3. 03

    Circuit Tracer and model steering tools

  4. 04

    Open-source interp-engine

Milestones

  1. 2023

    Neuronpedia created

    Started as a game for labeling neurons.

  2. 2024

    Refocus on researchers

    Shifted to supporting interpretability researchers.

  3. 2025

    Open-sourced

    Released the platform's code publicly on 31 March 2025.

  4. 2026

    Circuit Tracer

    Hosted the open circuit-tracing integration with Anthropic and partners.

Open source

GitHubgithub.com/hijohnnylin
Hugging Facehuggingface.co/neuronpedia

Notable releases

Openness

Apache-2.0 platform code with open dashboards data on Hugging Face.

People

  • Johnny LinFounder and lead

    Creator of Neuronpedia.

  • Curt TiggesScience lead, Decode Research

    Named by Anthropic as science lead of the Neuronpedia integration.

Listed from the lab's own pages and public profiles.

Contact and community

Via website.

Sources

Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.

Labs nearby