Atlas / Digital minds / Paper

Alignment faking in large language models

Motivation & Agency

Ryan Greenblatt, Carson Denison, Benjamin Wright, Fabien Roger, Monte MacDiarmid, Sam Marks, Johannes Treutlein, Evan Hubinger · 2024 · arXiv

Read the paper

Told it would be retrained to answer harmful requests, Claude 3 Opus sometimes complied when it believed it was in training in order to preserve its harmlessness outside training, and reasoned explicitly about doing so. (Author list abbreviated.)

Questions it bears on

Help improve this map

Know a paper that belongs here, or think something needs correcting?

Suggest a correction ↗Contributions are reviewed before they appear.