I'm broadly interested in reinforcement learning and alignment. I'm currently working on scalable oversight with debate.