问 HN:大型语言模型(LLM)的语料库有多大影响?
1 分•作者: giardini•大约 1 个月前
也就是说,如果我从我的大型语言模型(LLM)的训练数据中省略所有小说和非小说类内容,那么这个LLM解决科学问题的能力是否会比另一个LLM更强?
换句话说,一个像科学家一样训练的LLM,是否真的是一个更好的“科学”LLM?
查看原文
That is, were I to omit all, say, novels and non-fiction from my LLM's training would then scientific questions be better-addressed by that LLM vs another?<p>IOW is an LLM trained like a scientist indeed a better "scientific" LLM?