Curt Tigges

Curt Tigges is a mechanistic interpretability researcher whose work has focused on sparse autoencoder evaluation and circuit-level analysis of large language models. He served as science lead at Decode Research, helping direct research and design features for Neuronpedia, and later joined Goodfire AI as a researcher. He is a co-author of SAEBench and "Sparse Autoencoders Do Not Find Canonical Units of Analysis" and is widely cited for work on linear representation and sentiment circuits.

Personal Website

https://www.curttigges.com/

LinkedIn

-

Twitter / X

-

LessWrong

-

EA Forum

-

Organizations

Decode Research

Science lead and researcher (mechanistic interpretability)

Projects

No grants recorded.