Confessions can keep language models honest openai.com 4 points by xavierlint a year ago · 0 comments Reader PiP Save No comments yet.