Metaphor and Large Language Models: When Surface Features Matter More than Deep Understanding

TL;DR

This paper critically evaluates Large Language Models' ability to interpret metaphors, revealing their reliance on surface features rather than genuine understanding, through extensive experiments on diverse datasets and tasks.

Contribution

It provides a comprehensive, multi-dataset analysis of LLMs' metaphor interpretation, highlighting the influence of surface features over deep understanding, and emphasizes the need for realistic evaluation methods.

Findings

LLMs' performance depends more on lexical overlap and sentence length than on metaphorical content.

Surface features significantly influence LLMs' metaphor interpretation abilities.

Emergent metaphor understanding in LLMs is limited and context-dependent.

Abstract

This paper presents a comprehensive evaluation of the capabilities of Large Language Models (LLMs) in metaphor interpretation across multiple datasets, tasks, and prompt configurations. Although metaphor processing has gained significant attention in Natural Language Processing (NLP), previous research has been limited to single-dataset evaluations and specific task settings, often using artificially constructed data through lexical replacement. We address these limitations by conducting extensive experiments using diverse publicly available datasets with inference and metaphor annotations, focusing on Natural Language Inference (NLI) and Question Answering (QA) tasks. The results indicate that LLMs' performance is more influenced by features like lexical overlap and sentence length than by metaphorical content, demonstrating that any alleged emergent abilities of LLMs to understand metaphorical language are the result of a combination of surface-level features, in-context learning, and linguistic knowledge. This work provides critical insights into the current capabilities and limitations of LLMs in processing figurative language, highlighting the need for more realistic evaluation frameworks in metaphor interpretation tasks. Data and code are publicly available.