Saved in:
Bibliographic Details
Main Authors: Hwang, Jaehui, Han, Dongyoon, Yun, Sangdoo, Heo, Byeongho
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.17421
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908785809817600
author Hwang, Jaehui
Han, Dongyoon
Yun, Sangdoo
Heo, Byeongho
author_facet Hwang, Jaehui
Han, Dongyoon
Yun, Sangdoo
Heo, Byeongho
contents The emergence of discourse-like tokens such as "wait" and "therefore" in large language models (LLMs) has offered a unique window into their reasoning processes. However, systematic analyses of how such signals vary across training strategies and model scales remain lacking. In this paper, we analyze token-level signals through token probabilities across various models. We find that specific tokens strongly correlate with reasoning correctness, varying with training strategies while remaining stable across model scales. A closer look at the "wait" token in relation to answer probability demonstrates that models fine-tuned on small-scale datasets acquire reasoning ability through such signals but exploit them only partially. This work provides a systematic lens to observe and understand the dynamics of LLM reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2601_17421
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Oops, Wait: Token-Level Signals as a Lens into LLM Reasoning
Hwang, Jaehui
Han, Dongyoon
Yun, Sangdoo
Heo, Byeongho
Computation and Language
The emergence of discourse-like tokens such as "wait" and "therefore" in large language models (LLMs) has offered a unique window into their reasoning processes. However, systematic analyses of how such signals vary across training strategies and model scales remain lacking. In this paper, we analyze token-level signals through token probabilities across various models. We find that specific tokens strongly correlate with reasoning correctness, varying with training strategies while remaining stable across model scales. A closer look at the "wait" token in relation to answer probability demonstrates that models fine-tuned on small-scale datasets acquire reasoning ability through such signals but exploit them only partially. This work provides a systematic lens to observe and understand the dynamics of LLM reasoning.
title Oops, Wait: Token-Level Signals as a Lens into LLM Reasoning
topic Computation and Language
url https://arxiv.org/abs/2601.17421