Blog
Join our mailing list to stay up to date on our work.
Alignment Midtraining Cracks Under Pressure
Alignment
Obstacles to the scalable oversight of auto-alignment research
Alignment
Automated alignment runs are hard to study!
Alignment
Alejandro Aristizábal, Andrew Draganov, Aleksandr Bowkis, Cameron Holmes
Debate with Self-Play Best-of-N Optimization
Alignment
Dewi Gould, Sam Martin, Alejandro Aristizabal, Simon Marshall, Jacob Pfau
Your Model Organisms Might Be Fried
Alignment
Daniel Tan, Jonathan Bostock, Andrew Draganov, Angel Martinez, Sid Baines, David Africa
Learnings from starting an AI safety research team
Organisation
Andrew Draganov, Erin Robertson