Saved in:
Bibliographic Details
Main Authors: Kamath, Gaurav, Schuster, Sebastian, Vajjala, Sowmya, Reddy, Siva
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2404.04332
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909225079275520
author Kamath, Gaurav
Schuster, Sebastian
Vajjala, Sowmya
Reddy, Siva
author_facet Kamath, Gaurav
Schuster, Sebastian
Vajjala, Sowmya
Reddy, Siva
contents Sentences containing multiple semantic operators with overlapping scope often create ambiguities in interpretation, known as scope ambiguities. These ambiguities offer rich insights into the interaction between semantic structure and world knowledge in language processing. Despite this, there has been little research into how modern large language models treat them. In this paper, we investigate how different versions of certain autoregressive language models -- GPT-2, GPT-3/3.5, Llama 2 and GPT-4 -- treat scope ambiguous sentences, and compare this with human judgments. We introduce novel datasets that contain a joint total of almost 1,000 unique scope-ambiguous sentences, containing interactions between a range of semantic operators, and annotated for human judgments. Using these datasets, we find evidence that several models (i) are sensitive to the meaning ambiguity in these sentences, in a way that patterns well with human judgments, and (ii) can successfully identify human-preferred readings at a high level of accuracy (over 90% in some cases).
format Preprint
id arxiv_https___arxiv_org_abs_2404_04332
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scope Ambiguities in Large Language Models
Kamath, Gaurav
Schuster, Sebastian
Vajjala, Sowmya
Reddy, Siva
Computation and Language
Artificial Intelligence
Sentences containing multiple semantic operators with overlapping scope often create ambiguities in interpretation, known as scope ambiguities. These ambiguities offer rich insights into the interaction between semantic structure and world knowledge in language processing. Despite this, there has been little research into how modern large language models treat them. In this paper, we investigate how different versions of certain autoregressive language models -- GPT-2, GPT-3/3.5, Llama 2 and GPT-4 -- treat scope ambiguous sentences, and compare this with human judgments. We introduce novel datasets that contain a joint total of almost 1,000 unique scope-ambiguous sentences, containing interactions between a range of semantic operators, and annotated for human judgments. Using these datasets, we find evidence that several models (i) are sensitive to the meaning ambiguity in these sentences, in a way that patterns well with human judgments, and (ii) can successfully identify human-preferred readings at a high level of accuracy (over 90% in some cases).
title Scope Ambiguities in Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2404.04332