On the Tip of the Tongue: Analyzing Conceptual Representation in Large Language Models with Reverse-Dictionary Probe

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Ningyu, Zhang, Qi, Zhang, Menghan, Qian, Peng, Huang, Xuanjing
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913243703803904
author Xu, Ningyu
Zhang, Qi
Zhang, Menghan
Qian, Peng
Huang, Xuanjing
author_facet Xu, Ningyu
Zhang, Qi
Zhang, Menghan
Qian, Peng
Huang, Xuanjing
contents Probing and enhancing large language models' reasoning capacity remains a crucial open question. Here we re-purpose the reverse dictionary task as a case study to probe LLMs' capacity for conceptual inference. We use in-context learning to guide the models to generate the term for an object concept implied in a linguistic description. Models robustly achieve high accuracy in this task, and their representation space encodes information about object categories and fine-grained features. Further experiments suggest that the conceptual inference ability as probed by the reverse-dictionary task predicts model's general reasoning performance across multiple benchmarks, despite similar syntactic generalization behaviors across models. Explorative analyses suggest that prompting LLMs with description$\Rightarrow$word examples may induce generalization beyond surface-level differences in task construals and facilitate models on broader commonsense reasoning problems.
format Preprint
id arxiv_https___arxiv_org_abs_2402_14404
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On the Tip of the Tongue: Analyzing Conceptual Representation in Large Language Models with Reverse-Dictionary Probe
Xu, Ningyu
Zhang, Qi
Zhang, Menghan
Qian, Peng
Huang, Xuanjing
Computation and Language
Artificial Intelligence
Probing and enhancing large language models' reasoning capacity remains a crucial open question. Here we re-purpose the reverse dictionary task as a case study to probe LLMs' capacity for conceptual inference. We use in-context learning to guide the models to generate the term for an object concept implied in a linguistic description. Models robustly achieve high accuracy in this task, and their representation space encodes information about object categories and fine-grained features. Further experiments suggest that the conceptual inference ability as probed by the reverse-dictionary task predicts model's general reasoning performance across multiple benchmarks, despite similar syntactic generalization behaviors across models. Explorative analyses suggest that prompting LLMs with description$\Rightarrow$word examples may induce generalization beyond surface-level differences in task construals and facilitate models on broader commonsense reasoning problems.
title On the Tip of the Tongue: Analyzing Conceptual Representation in Large Language Models with Reverse-Dictionary Probe
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2402.14404