将推理令牌视为计算状态，而非文字叙述。
State over Tokens: Characterizing the Role of Reasoning Tokens
- 把推理过程看作模型外部化的计算状态，而非语言描述。
- 推理令牌虽非忠实解释，但能有效提升复杂任务表现。
- 适合研究大模型内在机制的学者，推动理解其真实推理过程。
大型语言模型（LLMs）在生成最终答案前会输出推理令牌以提升复杂任务的表现。尽管这些序列看似人类的思维过程，但实证研究表明它们并非模型实际推理过程的忠实解释。为弥合理论外观与功能之间的差距，我们提出「令牌上的状态」（State over Tokens, SoT）概念框架。SoT 将推理令牌重新定义为外部化的计算状态——是模型无状态生成周期中唯一持续传递的信息载体。这一视角解释了为何推理令牌无需作为可读文本仍能驱动正确推理，并揭示了此前被忽视的研究问题。我们认为，要真正理解 LLM 的推理过程，研究应超越将推理令牌作为文本阅读的传统方式，转而致力于解码其作为状态的实质。
原文摘要 · Abstract (English)
Large Language Models (LLMs) can generate reasoning tokens before their final answer to boost performance on complex tasks. While these sequences seem like human thought processes, empirical evidence reveals that they are not a faithful explanation of the model's actual reasoning process. To address this gap between appearance and function, we introduce the State over Tokens (SoT) conceptual framework. SoT reframes reasoning tokens not as a linguistic narrative, but as an externalized computational state -- the sole persistent information carrier across the model's stateless generation cycles. This explains how the tokens can drive correct reasoning without being a faithful explanation when read as text and surfaces previously overlooked research questions on these tokens. We argue that to truly understand the process that LLMs do, research must move beyond reading the reasoning tokens as text and focus on decoding them as state.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。