Logo Lanfrica
  • Accueil
  • Atlas
  • Analyses
  • Documentation
  • Sign in

© 2026 Lanfrica. Tous droits réservés. Tous les droits d'auteur des ressources affichées sur le site Web Lanfrica appartiennent aux détenteurs de droits d'auteur d'origine, sauf indication contraire explicite.

Multilinguality Does not Make Sense: Investigating Factors Behind Zero-Shot Transfer in Sense-Aware Tasks

Domaine:

natural language processing

Type de record:

papermodel
Créateur:
GowDub
Hôte:avatar
Cross-lingual transfer is central to modern NLP, enabling models to perform tasks in languages different from those they were trained on. A common assumption is that training on more languages improves zero-shot transfer. We test this on sense-aware tasks-polysemy and lexical semantic change-and find that multilinguality is not necessary for effective transfer. Our large-scale analysis across 28 languages reveals that other factors, such as differences in pretraining and fine-tuning data and evaluation artifacts, better explain the perceived benefits of multilinguality. We also release fine-tuned models and provide empirical baselines to support future research. While focused on two sense-aware tasks, our findings offer broader insights into cross-lingual transfer, especially for low-resource languages. accepted to EMNLP 2025 Main

Visit

arxiv.org

Tasks

transfer learning

Tags

Computation and Language

Similaires

29. Does Rian Venter make (Theological) Sense?How does the inclusion of multilingual intermediate tasks in XTREME-R affect zero-shot cross-lingual transfer performanceMultimodal Alignment Tasks and Zero-Shot Cross-Lingual Transfer in XLM-RDomain-Specific Intermediate Tasks and Zero-Shot Cross-Lingual Transfer in XNLIMultilingual Intermediate Tasks for Zero-Shot Cross-Lingual Transfer in XTREME-RScaling Intermediate Language Tasks for Zero-Shot Cross-Lingual Transfer in XTREME-R

29. Does Rian Venter make (Theological) Sense?

Engaging with the systematic-theologian Rian Venter in acknowledgement and celebration of his academ

How does the inclusion of multilingual intermediate tasks in XTREME-R affect zero-shot cross-lingual transfer performance

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuni

Multimodal Alignment Tasks and Zero-Shot Cross-Lingual Transfer in XLM-R

The introduction of pretrained cross-lingual language models brought decisive improvements to multil

Domain-Specific Intermediate Tasks and Zero-Shot Cross-Lingual Transfer in XNLI

Zero-shot cross-lingual transfer is a central task in multilingual NLP, allowing models trained in l

Multilingual Intermediate Tasks for Zero-Shot Cross-Lingual Transfer in XTREME-R

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuni

Scaling Intermediate Language Tasks for Zero-Shot Cross-Lingual Transfer in XTREME-R

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuni