Back to Directory
Dr. Logan Roberts
Stanford HAI
LessWrong
41
Papers
41
Posts
22
h-index
Links
Research Topics
InterpretabilityDeep LearningAI Risk
About
Dr. Logan Roberts is a researcher specializing in interpretability, deep learning, ai risk. They have published extensively in top-tier venues and are actively involved in the lesswrong community.
Scaling Monosemanticity: Extracting Interpretable Features from Large Language Models
NeurIPS2024156 citations
Co-authors: Neel Nanda, Jan Leike
Toward Understanding of Circuits in Transformers
ICML2023243 citations
Co-authors: Neel Nanda
Activation Patching: A Causal Lens on Neural Networks
ICLR2023189 citations
Co-authors: Paul Christiano