Ansh Radhakrishnan

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

arXiv 2024

Debating with More Persuasive LLMs Leads to More Truthful Answers

arXiv 2024

Question Decomposition Improves the Faithfulness of Model-Generated Reasoning

arXiv 2023

Ethan Perez

Samuel R. Bowman

Carson Denison

Evan Hubinger

Jan Brauner

Jared Kaplan

co-founder / Chief Science Officer

Kshitij Sachan

Newton Cheng

Nicholas Schiefer

Tamera Lanham