Memorization or Reasoning? Exploring the Idiom Understanding of LLMs

TL;DR

This paper investigates how large language models understand idioms across multiple languages, revealing that their processing involves both memorization and reasoning, especially for compositional idioms.

Contribution

Introduces MIDAS, a large-scale multilingual idiom dataset, and provides a comprehensive evaluation of LLMs' idiom understanding mechanisms.

Findings

LLMs rely on memorization and reasoning for idiom understanding

Performance varies across languages and idiom types

Contextual cues significantly influence LLMs' comprehension

Abstract

Idioms have long posed a challenge due to their unique linguistic properties, which set them apart from other common expressions. While recent studies have leveraged large language models (LLMs) to handle idioms across various tasks, e.g., idiom-containing sentence generation and idiomatic machine translation, little is known about the underlying mechanisms of idiom processing in LLMs, particularly in multilingual settings. To this end, we introduce MIDAS, a new large-scale dataset of idioms in six languages, each paired with its corresponding meaning. Leveraging this resource, we conduct a comprehensive evaluation of LLMs' idiom processing ability, identifying key factors that influence their performance. Our findings suggest that LLMs rely not only on memorization, but also adopt a hybrid approach that integrates contextual cues and reasoning, especially when processing compositional idioms. This implies that idiom understanding in LLMs emerges from an interplay between internal knowledge retrieval and reasoning-based inference.