Microstructures and Accuracy of Graph Recall by Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Yanbang, Cui, Hejie, Kleinberg, Jon
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913568215007232
author Wang, Yanbang
Cui, Hejie
Kleinberg, Jon
author_facet Wang, Yanbang
Cui, Hejie
Kleinberg, Jon
contents Graphs data is crucial for many applications, and much of it exists in the relations described in textual format. As a result, being able to accurately recall and encode a graph described in earlier text is a basic yet pivotal ability that LLMs need to demonstrate if they are to perform reasoning tasks that involve graph-structured information. Human performance at graph recall has been studied by cognitive scientists for decades, and has been found to often exhibit certain structural patterns of bias that align with human handling of social relationships. To date, however, we know little about how LLMs behave in analogous graph recall tasks: do their recalled graphs also exhibit certain biased patterns, and if so, how do they compare with humans and affect other graph reasoning tasks? In this work, we perform the first systematical study of graph recall by LLMs, investigating the accuracy and biased microstructures (local structural patterns) in their recall. We find that LLMs not only underperform often in graph recall, but also tend to favor more triangles and alternating 2-paths. Moreover, we find that more advanced LLMs have a striking dependence on the domain that a real-world graph comes from -- by yielding the best recall accuracy when the graph is narrated in a language style consistent with its original domain.
format Preprint
id arxiv_https___arxiv_org_abs_2402_11821
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Microstructures and Accuracy of Graph Recall by Large Language Models
Wang, Yanbang
Cui, Hejie
Kleinberg, Jon
Machine Learning
Computation and Language
Information Retrieval
Social and Information Networks
Graphs data is crucial for many applications, and much of it exists in the relations described in textual format. As a result, being able to accurately recall and encode a graph described in earlier text is a basic yet pivotal ability that LLMs need to demonstrate if they are to perform reasoning tasks that involve graph-structured information. Human performance at graph recall has been studied by cognitive scientists for decades, and has been found to often exhibit certain structural patterns of bias that align with human handling of social relationships. To date, however, we know little about how LLMs behave in analogous graph recall tasks: do their recalled graphs also exhibit certain biased patterns, and if so, how do they compare with humans and affect other graph reasoning tasks? In this work, we perform the first systematical study of graph recall by LLMs, investigating the accuracy and biased microstructures (local structural patterns) in their recall. We find that LLMs not only underperform often in graph recall, but also tend to favor more triangles and alternating 2-paths. Moreover, we find that more advanced LLMs have a striking dependence on the domain that a real-world graph comes from -- by yielding the best recall accuracy when the graph is narrated in a language style consistent with its original domain.
title Microstructures and Accuracy of Graph Recall by Large Language Models
topic Machine Learning
Computation and Language
Information Retrieval
Social and Information Networks
url https://arxiv.org/abs/2402.11821