aiqbench analysis
This study challenges the traditional view that idiom decomposability predicts syntactic flexibility, finding instead a weak negative correlation in language models. The research demonstrates that idiom representation stabilization relies on a combination of surprisal, frequency, and decomposability, with decomposability exerting the strongest training-dependent effect. These findings offer a more nuanced understanding of how distributional learning mechanisms interact with figurative language processing in neural systems.
Rethinking the Idiomaticity Decomposability Hypothesis: Evidence from Distributional Learning
arXiv:2606.03817v1 Announce Type: new Abstract: Idioms can be analysed in terms of their decomposability, the extent to which constituent meanings contribute to the figurative whole. Decomposability is thought to predict syntactic flexibility. Usage-based accounts instead attribute idiom behaviour to distributional experience, such as speaker familiarity and predictability. We examine these views using contextualised language models as controlled distributional learners. We propose a model-internal measure of decomposability and relate it to human ratings, syntactic flexibility, and predictability while tracking idiom learning during pretraining. Model-derived decomposability correlates weakly with human judgments and shows a small but consistent negative relationship with syntactic flexibility. Pretraining analyses show that stabilisation of idiom representations in models is not explained by frequency alone. Instead, surprisal, decomposability, and frequency all contribute, with decomposability showing the strongest training-dependent effect.
Abstract:Idioms can be analysed in terms of their decomposability, the extent to which constituent meanings contribute to the figurative whole. Decomposability is thought to predict syntactic flexibility. Usage-based accounts instead attribute idiom behaviour to distributional experience, such as speaker familiarity and predictability. We examine these views using contextualised language models as controlled distributional learners. We propose a model-internal measure of decomposability and relate it to human ratings, syntactic flexibility, and predictability while tracking idiom learning during pretraining. Model-derived decomposability correlates weakly with human judgments and shows a small but consistent negative relationship with syntactic flexibility. Pretraining analyses show that stabilisation of idiom representations in models is not explained by frequency alone. Instead, surprisal, decomposability, and frequency all contribute, with decomposability showing the strongest training-dependent effect.
Computation and Language (cs.CL)
arXiv:2606.03817 [cs.CL]
(or arXiv:2606.03817v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.03817 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 220 words
Related coverage
Zero Flux: Flow-Based Comparison of High-Dimensional Discrete Distributions
arXiv · Oct 2, 2026
Uncovering Latent Reasoning Strategies in Language Models
arXiv · Jul 21, 2026
Distributed Multi Robot Lunar Cargo Transportation via Phase Decomposed Reinforcement Learning
arXiv · Aug 25, 2026
Leveraging Discrete Function Decomposability for Scientific Design
arXiv · Sep 9, 2026