Attack Selection in Agentic AI Control Evaluations Meaningfully Decreases Safety

Add code
Jun 03, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: