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 publicly available: https://github.com/elisanchez-beep/metaphorLLM

In this paper, we present our approach to the AdMIRe (Advancing Multimodal Idiomaticity Representation) shared task, outlining the methodologies and strategies employed to tackle the challenges of idiomatic expressions in multimodal contexts. We discuss both successful and unsuccessful approaches, including the use of models of varying sizes and experiments involving zero- and few-shot learning. Our final submission, based on a zero-shot instruction-following vision-and-language model (VLM), achieved 13th place for the English test set and 1st place for the Portuguese test set on the preliminary leaderboard.We investigate the performance of open VLMs in this task, demonstrating that both large language models (LLMs) and VLMs exhibit strong capabilities in identifying idiomatic expressions. However, we also identify significant limitations in both model types, including instability and a tendency to generate hallucinated content, which raises concerns about their reliability in interpreting figurative language. Our findings emphasize the need for further advancements in multimodal models to improve their robustness and mitigate these issues.

The lack of wide coverage datasets annotated with everyday metaphorical expressions for languages other than English is striking. This means that most research on supervised metaphor detection has been published only for that language. In order to address this issue, this work presents the first corpus annotated with naturally occurring metaphors in Spanish large enough to develop systems to perform metaphor detection. The presented dataset, CoMeta, includes texts from various domains, namely, news, political discourse, Wikipedia and reviews. In order to label CoMeta, we apply the MIPVU method, the guidelines most commonly used to systematically annotate metaphor on real data. We use our newly created dataset to provide competitive baselines by fine-tuning several multilingual and monolingual state-of-the-art large language models. Furthermore, by leveraging the existing VUAM English data in addition to CoMeta, we present the, to the best of our knowledge, first cross-lingual experiments on supervised metaphor detection. Finally, we perform a detailed error analysis that explores the seemingly high transfer of everyday metaphor across these two languages and datasets.