Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Qin, Tian, Chen, Yuhan, Wang, Zhiwei, Xu, Zhi-Qin John
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2509.23178
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912611611705344
author Qin, Tian
Chen, Yuhan
Wang, Zhiwei
Xu, Zhi-Qin John
author_facet Qin, Tian
Chen, Yuhan
Wang, Zhiwei
Xu, Zhi-Qin John
contents Transformers are able to perform reasoning tasks, however the intrinsic mechanism remains widely open. In this paper we propose a set of information propagation rules based on Transformers and utilize symbolic reasoning tasks to theoretically analyze the limit reasoning steps. We show that the limit number of reasoning steps is between $O(3^{L-1})$ and $O(2^{L-1})$ for a model with $L$ attention layers in a single-pass.
format Preprint
id arxiv_https___arxiv_org_abs_2509_23178
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Limit Analysis for Symbolic Multi-step Reasoning Tasks with Information Propagation Rules Based on Transformers
Qin, Tian
Chen, Yuhan
Wang, Zhiwei
Xu, Zhi-Qin John
Artificial Intelligence
Transformers are able to perform reasoning tasks, however the intrinsic mechanism remains widely open. In this paper we propose a set of information propagation rules based on Transformers and utilize symbolic reasoning tasks to theoretically analyze the limit reasoning steps. We show that the limit number of reasoning steps is between $O(3^{L-1})$ and $O(2^{L-1})$ for a model with $L$ attention layers in a single-pass.
title Limit Analysis for Symbolic Multi-step Reasoning Tasks with Information Propagation Rules Based on Transformers
topic Artificial Intelligence
url https://arxiv.org/abs/2509.23178