ParentFull threadjstx1·How well do these papers replicate? In some of experiments with GPT-4 I've seen chain-of-thought style prompting make answers noticeably worse than plainly asking a question once.View on HN