Question: Scalable Oversight Protocols for Superhuman AI

Debate protocol accuracy on hard NLP tasks

Collecting findings

0/5 publication-ready

Findings metrics

—

total

—

verified

—

pending

—

rejected

Research