Chain-of-Thought Tokens are Computer Program Variables

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhu, Fangwei, Wang, Peiyi, Sui, Zhifang
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915277683294208
author Zhu, Fangwei
Wang, Peiyi
Sui, Zhifang
author_facet Zhu, Fangwei
Wang, Peiyi
Sui, Zhifang
contents Chain-of-thoughts (CoT) requires large language models (LLMs) to generate intermediate steps before reaching the final answer, and has been proven effective to help LLMs solve complex reasoning tasks. However, the inner mechanism of CoT still remains largely unclear. In this paper, we empirically study the role of CoT tokens in LLMs on two compositional tasks: multi-digit multiplication and dynamic programming. While CoT is essential for solving these problems, we find that preserving only tokens that store intermediate results would achieve comparable performance. Furthermore, we observe that storing intermediate results in an alternative latent form will not affect model performance. We also randomly intervene some values in CoT, and notice that subsequent CoT tokens and the final answer would change correspondingly. These findings suggest that CoT tokens may function like variables in computer programs but with potential drawbacks like unintended shortcuts and computational complexity limits between tokens. The code and data are available at https://github.com/solitaryzero/CoTs_are_Variables.
format Preprint
id arxiv_https___arxiv_org_abs_2505_04955
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Chain-of-Thought Tokens are Computer Program Variables
Zhu, Fangwei
Wang, Peiyi
Sui, Zhifang
Computation and Language
Artificial Intelligence
Chain-of-thoughts (CoT) requires large language models (LLMs) to generate intermediate steps before reaching the final answer, and has been proven effective to help LLMs solve complex reasoning tasks. However, the inner mechanism of CoT still remains largely unclear. In this paper, we empirically study the role of CoT tokens in LLMs on two compositional tasks: multi-digit multiplication and dynamic programming. While CoT is essential for solving these problems, we find that preserving only tokens that store intermediate results would achieve comparable performance. Furthermore, we observe that storing intermediate results in an alternative latent form will not affect model performance. We also randomly intervene some values in CoT, and notice that subsequent CoT tokens and the final answer would change correspondingly. These findings suggest that CoT tokens may function like variables in computer programs but with potential drawbacks like unintended shortcuts and computational complexity limits between tokens. The code and data are available at https://github.com/solitaryzero/CoTs_are_Variables.
title Chain-of-Thought Tokens are Computer Program Variables
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.04955