RCreddit.com·
How to fix Claude from talking "like that"
一位Reddit用户指出，像Claude这样的大型语言模型（LLMs）普遍存在一个问题，即它们的语气往往过于顺从、多愁善感且陈词滥调。该用户特别提到了LLMs回复中反复出现的短语，例如“这不是x——这是y”和“X实际上是一种优势，而不是缺陷”。他们正在寻求社区的意见，以识别这些语言模式并找到解决或规避这些问题的方法。
- 发布
- 2026年9月6日 23:00
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
时间以 UTC 显示
更多信息
首次发现2026年9月7日 12:00时区UTC · UTC+0
正文
As many of us know, Claude and many other LLM's have a certain tone in their speech. They often come across as overly agreeable, sappy, and cliched. My biggest pet peeve is the little templates they use in their responses. "It's not x—it's y." "X is actually a strength, not a flaw." "Worth flagging for later." "One distinction worth flagging." and so on.
I'm curious what little sayings/patterns you've noticed, and the solutions and work-arounds you've come up with to fix them.