Language Models Do Hard Arithmetic Tasks Easily and Hardly Do Easy Arithmetic Tasks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gambardella, Andrew, Iwasawa, Yusuke, Matsuo, Yutaka
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916273893408768
author Gambardella, Andrew
Iwasawa, Yusuke
Matsuo, Yutaka
author_facet Gambardella, Andrew
Iwasawa, Yusuke
Matsuo, Yutaka
contents The ability (and inability) of large language models (LLMs) to perform arithmetic tasks has been the subject of much theoretical and practical debate. We show that LLMs are frequently able to correctly and confidently predict the first digit of n-digit by m-digit multiplication tasks without using chain of thought reasoning, despite these tasks require compounding operations to solve. Simultaneously, LLMs in practice often fail to correctly or confidently predict the last digit of an n-digit by m-digit multiplication, a task equivalent to 1-digit by 1-digit multiplication which can be easily learned or memorized. We show that the latter task can be solved more robustly when the LLM is conditioned on all of the correct higher-order digits, which on average increases the confidence of the correct last digit on 5-digit by 5-digit multiplication tasks using Llama 2-13B by over 230% (0.13 to 0.43) and Mistral-7B by 150% (0.22 to 0.55).
format Preprint
id arxiv_https___arxiv_org_abs_2406_02356
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Language Models Do Hard Arithmetic Tasks Easily and Hardly Do Easy Arithmetic Tasks
Gambardella, Andrew
Iwasawa, Yusuke
Matsuo, Yutaka
Machine Learning
Artificial Intelligence
Computation and Language
The ability (and inability) of large language models (LLMs) to perform arithmetic tasks has been the subject of much theoretical and practical debate. We show that LLMs are frequently able to correctly and confidently predict the first digit of n-digit by m-digit multiplication tasks without using chain of thought reasoning, despite these tasks require compounding operations to solve. Simultaneously, LLMs in practice often fail to correctly or confidently predict the last digit of an n-digit by m-digit multiplication, a task equivalent to 1-digit by 1-digit multiplication which can be easily learned or memorized. We show that the latter task can be solved more robustly when the LLM is conditioned on all of the correct higher-order digits, which on average increases the confidence of the correct last digit on 5-digit by 5-digit multiplication tasks using Llama 2-13B by over 230% (0.13 to 0.43) and Mistral-7B by 150% (0.22 to 0.55).
title Language Models Do Hard Arithmetic Tasks Easily and Hardly Do Easy Arithmetic Tasks
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2406.02356