请问HN:大型语言模型的语料库差异有多大?

1作者: giardini大约 1 个月前原帖
也就是说,如果我在我的大型语言模型(LLM)的训练中省略所有小说和非虚构类作品,那么这个LLM在处理科学问题时是否会比其他模型更有效?换句话说,一个像科学家一样训练的LLM是否真的会成为更“科学”的LLM?
查看原文
That is, were I to omit all, say, novels and non-fiction from my LLM&#x27;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 &quot;scientific&quot; LLM?