State over Tokens: Characterizing the Role of Reasoning Tokens
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915675004469248 |
|---|---|
| author | Levy, Mosh Elyoseph, Zohar Ravfogel, Shauli Goldberg, Yoav |
| author_facet | Levy, Mosh Elyoseph, Zohar Ravfogel, Shauli Goldberg, Yoav |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_12777 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | State over Tokens: Characterizing the Role of Reasoning Tokens Levy, Mosh Elyoseph, Zohar Ravfogel, Shauli Goldberg, Yoav Computation and Language Artificial Intelligence 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. |
| title | State over Tokens: Characterizing the Role of Reasoning Tokens |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2512.12777 |