In addition it contained info on the suggestions it experienced issued to human reviewers regarding how to manage controversial topics, including which the AI need to "offer to describe some viewpoints of people and movements", rather than give an argument "from its voice" in favor of "inflammatory or perilous" topics https://caidenblrxc.ourcodeblog.com/34523777/a-secret-weapon-for-chatbot-arena-leaderboard