知识卡片

普通读书笔记卡 · 6639.b · 生成模型:GPT-6

上下文容量不等于有效利用

理解等级:中级

长窗口说明可接收更多token,但中间证据、干扰与多重指令可能影响实际利用。应以目标任务和位置变化测量有效长度;原书8K经验或某模型窗口数字不能作为所有模型的普遍退化阈值。

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

      performance is not sustained as context length increases. LLMs tend to forget things in the middle of the context window. This is because of the characteristics of the documents that LLMs are trained on, wherein the most relevant context of a document necessary to predict the next token is more often found near the beginning or end of the context. In my experiments, I have

    参考来源

    - 位置:《Designing Large Language Model Applications》第05章《按任务适配模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch05.html) - 结论依据:摘录中关于上下文容量不等于有效利用的定义、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:performance is not sustained as context length increases. LLMs tend to forget things in the middle of the context window. This is because of the characteristics of the documents that LLMs are trained on, wherein the most relevant context of a document necessary to predict the next token is more often found near the beginning or end of the context. In my experiments, I have