Saved in:
Bibliographic Details
Main Authors: Amiri, Alireza, Huang, Xinting, Rofin, Mark, Hahn, Michael
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.02393
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913938669568000
author Amiri, Alireza
Huang, Xinting
Rofin, Mark
Hahn, Michael
author_facet Amiri, Alireza
Huang, Xinting
Rofin, Mark
Hahn, Michael
contents Chain-of-thought reasoning and scratchpads have emerged as critical tools for enhancing the computational capabilities of transformers. While theoretical results show that polynomial-length scratchpads can extend transformers' expressivity from $TC^0$ to $PTIME$, their required length remains poorly understood. Empirical evidence even suggests that transformers need scratchpads even for many problems in $TC^0$, such as Parity or Multiplication, challenging optimistic bounds derived from circuit complexity. In this work, we initiate the study of systematic lower bounds for the number of chain-of-thought steps across different algorithmic problems, in the hard-attention regime. We study a variety of algorithmic problems, and provide bounds that are tight up to logarithmic factors. Overall, these results contribute to emerging understanding of the power and limitations of chain-of-thought reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2502_02393
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lower Bounds for Chain-of-Thought Reasoning in Hard-Attention Transformers
Amiri, Alireza
Huang, Xinting
Rofin, Mark
Hahn, Michael
Machine Learning
Computational Complexity
Chain-of-thought reasoning and scratchpads have emerged as critical tools for enhancing the computational capabilities of transformers. While theoretical results show that polynomial-length scratchpads can extend transformers' expressivity from $TC^0$ to $PTIME$, their required length remains poorly understood. Empirical evidence even suggests that transformers need scratchpads even for many problems in $TC^0$, such as Parity or Multiplication, challenging optimistic bounds derived from circuit complexity. In this work, we initiate the study of systematic lower bounds for the number of chain-of-thought steps across different algorithmic problems, in the hard-attention regime. We study a variety of algorithmic problems, and provide bounds that are tight up to logarithmic factors. Overall, these results contribute to emerging understanding of the power and limitations of chain-of-thought reasoning.
title Lower Bounds for Chain-of-Thought Reasoning in Hard-Attention Transformers
topic Machine Learning
Computational Complexity
url https://arxiv.org/abs/2502.02393