Enregistré dans:
Détails bibliographiques
Auteurs principaux: Trott, Sean, Rivière, Pamela D.
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:https://arxiv.org/abs/2602.05035
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866912878998585344
author Trott, Sean
Rivière, Pamela D.
author_facet Trott, Sean
Rivière, Pamela D.
contents Multilingual language models (LMs) sometimes under-perform their monolingual counterparts, possibly due to capacity limitations. We quantify this ``multilingual penalty'' for lexical disambiguation--a task requiring precise semantic representations and contextualization mechanisms--using controlled datasets of human relatedness judgments for ambiguous words in both English and Spanish. Comparing monolingual and multilingual LMs from the same families, we find consistently reduced performance in multilingual LMs. We then explore three potential capacity constraints: representational (reduced embedding isotropy), attentional (reduced attention to disambiguating cues), and vocabulary-related (increased multi-token segmentation). Multilingual LMs show some evidence of all three limitations; moreover, these factors statistically account for the variance formerly attributed to a model's multilingual status. These findings suggest both that multilingual LMs do suffer from multiple capacity constraints, and that these constraints correlate with reduced disambiguation performance.
format Preprint
id arxiv_https___arxiv_org_abs_2602_05035
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Capacity Constraints and the Multilingual Penalty for Lexical Disambiguation
Trott, Sean
Rivière, Pamela D.
Computation and Language
Multilingual language models (LMs) sometimes under-perform their monolingual counterparts, possibly due to capacity limitations. We quantify this ``multilingual penalty'' for lexical disambiguation--a task requiring precise semantic representations and contextualization mechanisms--using controlled datasets of human relatedness judgments for ambiguous words in both English and Spanish. Comparing monolingual and multilingual LMs from the same families, we find consistently reduced performance in multilingual LMs. We then explore three potential capacity constraints: representational (reduced embedding isotropy), attentional (reduced attention to disambiguating cues), and vocabulary-related (increased multi-token segmentation). Multilingual LMs show some evidence of all three limitations; moreover, these factors statistically account for the variance formerly attributed to a model's multilingual status. These findings suggest both that multilingual LMs do suffer from multiple capacity constraints, and that these constraints correlate with reduced disambiguation performance.
title Capacity Constraints and the Multilingual Penalty for Lexical Disambiguation
topic Computation and Language
url https://arxiv.org/abs/2602.05035