Scalable Oversight for Superhuman AI via Recursive Self-Critiquing

Open in new window