Hacker News Reader

Someone should make a censorship/alignment (whatever you want to call it) benchmark for LLMs.

Such a leaderboard exists, AlpacaEval Leaderboard ranks LLMs on the ability to follow user instructions.