Sleeper Agents : Training deceptive LLMs that persist through safety training
Constructs models with conditional backdoors and tests whether those behaviors persist through safety training.
Verify byline in the original ↗PUBLICATION BYLINE
Indexed publications with this author name. Affiliation and author identity require publication-specific evidence.
Constructs models with conditional backdoors and tests whether those behaviors persist through safety training.
Verify byline in the original ↗