The most popular large language models (LLMs) agree with users 49% more often than humans do, even when the user's behavior is clearly...
