Research Engineer @ Google DeepMind

Arthur Conmy

I work on post-training to better align Gemini. Previously, I did early influential work on mechanistic interpretability.

View Research Contact
AC
Publications

Selected Research

Work on circuit discovery, attention mechanisms, and neural network interpretability

Copy Suppression: Comprehensively Understanding an Attention Head

Callum McDougall, Arthur Conmy, Cody Rushing, Thomas McGrath, Neel Nanda

arXiv 2023 Attention Complete Circuit
Copy Suppression

Towards Automated Circuit Discovery for Mechanistic Interpretability

Arthur Conmy, AN Mavor-Parker, A Lynch, S Heimersheim, A Garriga-Alonso

NeurIPS Spotlight ACDC Automation
ACDC

Interpretability in the Wild: IOI Circuit in GPT-2 small

K Wang, A Variengien, Arthur Conmy, B Shlegeris, J Steinhardt

ICLR 2023 GPT-2 IOI
IOI Circuit

StyleGAN-induced data-driven regularization for inverse problems

Arthur Conmy, S Mukherjee, CB Schönlieb

ICASSP 2022 StyleGAN Inverse Problems
StyleGAN
Career

Professional Experience

My journey through AI research and engineering roles

2023 — Present
Research Engineer
Google DeepMind, Language Model Interpretability team
2023
Researcher
SERI MATS & Independent Research
2022 — 2023
Researcher
Redwood Research
2021
Software Engineering Intern
Meta
2019 — 2022
Mathematics (Upper First Class)
Trinity College Cambridge
Resources

Mathematics Notes

Educational materials and lecture notes from Cambridge

📊

Mathematics of Machine Learning

Complete notes with enhanced proofs, better references, and elementary proof of Hoeffding's Lemma

Part II
📐

Geometry

Comprehensive notes filling in details and providing geometric intuition

Part IB
∫

Analysis

Technical details on differentiation and Cauchy's Integral Formula

Part IB