AlignFreeze: Navigating the Impact of Realignment on the Layers of Multilingual Models Across Diverse Languages

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Bakos, Steve, Gaschi, Félix, Guzmán, David, More, Riddhi, Li, Kelly Chutong, Lee, En-Shiun Annie
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910834749341696
author Bakos, Steve
Gaschi, Félix
Guzmán, David
More, Riddhi
Li, Kelly Chutong
Lee, En-Shiun Annie
author_facet Bakos, Steve
Gaschi, Félix
Guzmán, David
More, Riddhi
Li, Kelly Chutong
Lee, En-Shiun Annie
contents Realignment techniques are often employed to enhance cross-lingual transfer in multilingual language models, still, they can sometimes degrade performance in languages that differ significantly from the fine-tuned source language. This paper introduces AlignFreeze, a method that freezes either the layers' lower half or upper half during realignment. Through controlled experiments on 4 tasks, 3 models, and in 35 languages, we find that realignment affects all the layers but can be the most detrimental to the lower ones. Freezing the lower layers can prevent performance degradation. Particularly, AlignFreeze improves Part-of-Speech (PoS) tagging performances in languages where full realignment fails: with XLM-R, it provides improvements of more than one standard deviation in accuracy in seven more languages than full realignment.
format Preprint
id arxiv_https___arxiv_org_abs_2502_12959
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AlignFreeze: Navigating the Impact of Realignment on the Layers of Multilingual Models Across Diverse Languages
Bakos, Steve
Gaschi, Félix
Guzmán, David
More, Riddhi
Li, Kelly Chutong
Lee, En-Shiun Annie
Computation and Language
Artificial Intelligence
Realignment techniques are often employed to enhance cross-lingual transfer in multilingual language models, still, they can sometimes degrade performance in languages that differ significantly from the fine-tuned source language. This paper introduces AlignFreeze, a method that freezes either the layers' lower half or upper half during realignment. Through controlled experiments on 4 tasks, 3 models, and in 35 languages, we find that realignment affects all the layers but can be the most detrimental to the lower ones. Freezing the lower layers can prevent performance degradation. Particularly, AlignFreeze improves Part-of-Speech (PoS) tagging performances in languages where full realignment fails: with XLM-R, it provides improvements of more than one standard deviation in accuracy in seven more languages than full realignment.
title AlignFreeze: Navigating the Impact of Realignment on the Layers of Multilingual Models Across Diverse Languages
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2502.12959