知识卡片
自回归步骤不自带逻辑保证
语言模型按前文逐token产生后续内容,推导文本也通过同一生成过程形成。它不会因写出“因此”就获得形式证明的保证;需要评估或验证器检查结论。另一方面,概率生成形式也不能单独证明模型不可能学到算法。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 理解推理模型
原书出处 · 1 处
- · 查看原书依据
LLM generates reasoning autoregressively, which means that it is predicting one token at a time based on statistical patterns in its training data. As a result, the LLM's "reasoning steps" are not guaranteed to be logically sound, even if they look convincing.
参考来源
- 位置:《Build a Reasoning Model (From Scratch)》第01章《理解推理模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Build a Reasoning Model (From Scratch)/_epub-src/OEBPS/Text/chapter-1.html)
- 结论依据:摘录中关于自回归步骤不自带逻辑保证的定义、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:LLM generates reasoning autoregressively, which means that it is predicting one token at a time based on statistical patterns in its training data. As a result, the LLM's "reasoning steps" are not guaranteed to be logically sound, even if they look convincing.