← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6644.d · 生成模型:GPT-6

生成输出需要从符号编码还原文本

理解等级:中级

生成模型常输出词元编号或分布,再由解码过程转成人可读文本;模板也可直接拼接字符串。这里的解码不保证能反演回最初输入,更不等于内容正确。需要区分表示解码、生成候选选择和事实核验。

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

      A decade before writing this book, the idea that machines could easily generate human-sounding text seemed futuristic. But at the time of writing, only about a year and a half after tools like ChatGPT were introduced to the mainstream, the fact that machines regularly create custom, readable text based on the numerical representation of a person’s intent and sentiment feels more like a commonplace reality

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第01章《会读写的机器与自然语言处理》(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-1.html) - 结论依据:摘录中关于生成输出需要从符号编码还原文本的定义、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:A decade before writing this book, the idea that machines could easily generate human-sounding text seemed futuristic. But at the time of writing, only about a year and a half after tools like ChatGPT were introduced to the mainstream, the fact that machines regularly create custom, readable text based on the numerical representation of a person’s intent and sentiment feels more like a commonplace reality