Buck comments on Meta-level adversarial evaluation of oversight techniques might allow robust measurement of their adequacy

[Buck]

I’m interested in both of those.