长时间智能体任务中的语言质量漂移
现在最令人恼火的事情是,在使用大语言模型进行长时间运行的智能体任务时,最烦人的不是编码、错误或幻觉,而是随着任务的进行,它们的语言由于漂移而变得越来越糟糕,这真是讽刺。
这就在你的名字里了!赶紧写得更好点吧!
是的,我有 voice.md 文件。是的,我有智能体作为读者进行最终审阅。是的,我有其他智能体专门查找 LLM 语言。是的,我有来自模型家族的不同模型,采用不同方法审阅面向用户的文本。这仍然不够。
对照原文
It is ironic that the thing that is now most annoying about long-running agentic tasks with Large Language Models isn't coding or errors or hallucinations, but the fact that their language gets worse due to drift as a task goes on Its in your the name! Just write better already!
Yes, I have voice.md files. Yes, I have agents doing final passes as a reader. Yes, I have other agents specifically looking for LLM language. Yes, I have different models from the model family with different approaches going over user facing text. It still isn't enough.