Guardado en:
Detalles Bibliográficos
Autores principales: Fekete, Marcell, Bjerva, Johannes
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2410.13396
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866929730929819648
author Fekete, Marcell
Bjerva, Johannes
author_facet Fekete, Marcell
Bjerva, Johannes
contents Understanding how linguistic knowledge is encoded in language models is crucial for improving their generalisation capabilities. In this paper, we investigate the processing of morphosyntactic phenomena, by leveraging a recently proposed method for probing language models via Shapley Head Values (SHVs). Using the English language BLiMP dataset, we test our approach on two widely used models, BERT and RoBERTa, and compare how linguistic constructions such as anaphor agreement and filler-gap dependencies are handled. Through quantitative pruning and qualitative clustering analysis, we demonstrate that attention heads responsible for processing related linguistic phenomena cluster together. Our results show that SHV-based attributions reveal distinct patterns across both models, providing insights into how language models organize and process linguistic information. These findings support the hypothesis that language models learn subnetworks corresponding to linguistic theory, with potential implications for cross-linguistic model analysis and interpretability in Natural Language Processing (NLP).
format Preprint
id arxiv_https___arxiv_org_abs_2410_13396
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Linguistically Grounded Analysis of Language Models using Shapley Head Values
Fekete, Marcell
Bjerva, Johannes
Computation and Language
Understanding how linguistic knowledge is encoded in language models is crucial for improving their generalisation capabilities. In this paper, we investigate the processing of morphosyntactic phenomena, by leveraging a recently proposed method for probing language models via Shapley Head Values (SHVs). Using the English language BLiMP dataset, we test our approach on two widely used models, BERT and RoBERTa, and compare how linguistic constructions such as anaphor agreement and filler-gap dependencies are handled. Through quantitative pruning and qualitative clustering analysis, we demonstrate that attention heads responsible for processing related linguistic phenomena cluster together. Our results show that SHV-based attributions reveal distinct patterns across both models, providing insights into how language models organize and process linguistic information. These findings support the hypothesis that language models learn subnetworks corresponding to linguistic theory, with potential implications for cross-linguistic model analysis and interpretability in Natural Language Processing (NLP).
title Linguistically Grounded Analysis of Language Models using Shapley Head Values
topic Computation and Language
url https://arxiv.org/abs/2410.13396