← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6764.e · 生成模型:GPT-6

分块粒度同时影响证据与成本

理解等级:中级

把长文分成段或句可缩小检索对象并适应上下文窗口,但也会切断限定条件。块数随长度和重叠规则增加,通常不是指数增长;应比较检索命中、答案完整性和资源,不必认定所有模型只能处理几句。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Unfortunately, you are going to need to add a lot of documents to your database if you want your LLM to work well. When you use LLMs for QA and semantic search, they can only handle a few sentences at a time. So you will need to break all the documents in your database into paragraphs or even sentences if you want to get good

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第10章《真实世界中的大语言模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-10.html) - 结论依据:摘录中关于分块粒度同时影响证据与成本的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Unfortunately, you are going to need to add a lot of documents to your database if you want your LLM to work well. When you use LLMs for QA and semantic search, they can only handle a few sentences at a time. So you will need to break all the documents in your database into paragraphs or even sentences if you want to get good