← 返回刚才阅读的位置

知识卡片

结构图卡 · 6762.g · 生成模型:GPT-6

生成过程由分布与选择循环组成

理解等级:中级

结构图:

flowchart LR
 A[匹配分词器与可见前缀] --> B[因果模型输出下一词logits]
 B --> C[按解码策略处理候选]
 C --> D[选择token并更新前缀]
 D --> E{EOS或预算耗尽}
 E -->|否| B
 E -->|是| F[解码并检查任务结果]

选择策略可以是贪心、束搜索或采样,模型分数与事实支持不能合成同一个指标。

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

      Now that you have learned about the dangers of using LLMs and how to mitigate them, it is time to focus on how they work as well as they do—and even generate some text yourself. So how does a generative model create new text? Under the hood, a language model is a conditional probability distribution function. A conditional distribution function gives you the probabilities of

    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:摘录中关于生成过程由分布与选择循环组成的定义、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:Now that you have learned about the dangers of using LLMs and how to mitigate them, it is time to focus on how they work as well as they do—and even generate some text yourself. So how does a generative model create new text? Under the hood, a language model is a conditional probability distribution function. A conditional distribution function gives you the probabilities of