Question: Constitutional AI and RLHF Alignment Comparison

RLHF helpfulness-safety Pareto frontier metric

In audit

1/5 publication-ready

Findings metrics

—

total

—

verified

—

pending

—

rejected

Research