Nature Machine Intelligence：将大语言模型推理 token 刻画为 token 之上的状态
Characterizing the role of reasoning tokens in large language models as state over tokens
摘要
Nature Machine Intelligence 在线发表 Levy 等人的 Perspective，提出将大语言模型的推理 token 刻画为"token 之上的状态"，用于考察这些推理 token 的作用。文章指出，大语言模型在"推理"一个问题时，实际解题方式可能与其推理文本所呈现的并不一致。
摘要由模型根据原文撰写，以原文为准。
正文 · 原文
Nature Machine Intelligence, Published online: 07 October 2026; doi:10.1038/s42256-026-01303-y
When large language models ‘reason’ about a problem, they might still solve it differently from what their reasoning transcript suggest. In this Perspective, Levy et al. describe how one can start investigating the role of these reasoning tokens.
注意转载授权 · 阅读原文
来源：Nature Machine Intelligence · nature.com
引用原文
Mosh Levy. Characterizing the role of reasoning tokens in large language models as state over tokens[EB/OL]. (2026-10-07)[访问日期]. https://www.nature.com/articles/s42256-026-01303-y.