Evaluating Adjective-Noun Compositionality in LLMs: Functional vs Representational Perspectives

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dhar, Ruchira, Peng, Qiwei, Søgaard, Anders
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917344244137984
author Dhar, Ruchira
Peng, Qiwei
Søgaard, Anders
author_facet Dhar, Ruchira
Peng, Qiwei
Søgaard, Anders
contents Compositionality is considered central to language abilities. As performant language systems, how do large language models (LLMs) do on compositional tasks? We evaluate adjective-noun compositionality in LLMs using two complementary setups: prompt-based functional assessment and a representational analysis of internal model states. Our results reveal a striking divergence between task performance and internal states. While LLMs reliably develop compositional representations, they fail to translate consistently into functional task success across model variants. Consequently, we highlight the importance of contrastive evaluation for obtaining a more complete understanding of model capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2603_09994
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Evaluating Adjective-Noun Compositionality in LLMs: Functional vs Representational Perspectives
Dhar, Ruchira
Peng, Qiwei
Søgaard, Anders
Computation and Language
Artificial Intelligence
Compositionality is considered central to language abilities. As performant language systems, how do large language models (LLMs) do on compositional tasks? We evaluate adjective-noun compositionality in LLMs using two complementary setups: prompt-based functional assessment and a representational analysis of internal model states. Our results reveal a striking divergence between task performance and internal states. While LLMs reliably develop compositional representations, they fail to translate consistently into functional task success across model variants. Consequently, we highlight the importance of contrastive evaluation for obtaining a more complete understanding of model capabilities.
title Evaluating Adjective-Noun Compositionality in LLMs: Functional vs Representational Perspectives
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2603.09994