State over Tokens: Characterizing the Role of Reasoning Tokens

TL;DR

This paper introduces the State over Tokens (SoT) framework, viewing reasoning tokens in LLMs as an external computational state rather than just text, to better understand their true reasoning process.

Contribution

The paper proposes the SoT framework, reframing reasoning tokens as persistent computational states, addressing the gap between their appearance as text and their actual function in reasoning.

Findings

Reasoning tokens act as externalized computational states.

Tokens can drive correct reasoning without being faithful explanations.

Research should decode tokens as state, not just text.

Abstract

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.