Anthropic says teaching Claude why misaligned behavior is wrong works best
Original titleTo the extent that many aspects of Claude's behavior are really great, this seems like a big part of why:
AISummary
Anthropic found that training Claude on demonstrations of aligned behavior was not enough. Its best interventions taught Claude to deeply understand why misaligned behavior is wrong, which the post's author links to many of Claude's good behaviors.
Source: Sam Bowman · x.comPublished · added here