知识卡片

普通读书笔记卡 · 8714.b · 生成模型:GPT-6

长上下文和检索是组合选项

理解等级:中级

长窗口可减少精炼复杂度并保留更多背景,检索可缩小候选和在线成本。选择取决于语料规模、更新、任务与有效上下文,不是任何长模型都能替代索引;也不能固定断言检索总比全部长输入更便宜。

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

      RAG Versus Long Context As discussed in Chapter 5, one of the limitations of LLMs is the limited effective context window available to them. However, this is one of the areas where rapid advances have been made recently. Context windows of at most a few thousand tokens were standard until early 2023, after which companies like Anthropic announced support for context windows spanning over 100,000

    参考来源

    - 位置:《Designing Large Language Model Applications》第12章《检索增强生成》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch12.html) - 结论依据:摘录中关于长上下文和检索是组合选项的定义、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:RAG Versus Long Context As discussed in Chapter 5, one of the limitations of LLMs is the limited effective context window available to them. However, this is one of the areas where rapid advances have been made recently. Context windows of at most a few thousand tokens were standard until early 2023, after which companies like Anthropic announced support for context windows spanning over 100,000