Advanced search
1 file | 453.97 KB Add to list

Lost in activations : a neuron-level analysis of encoders for cross-lingual emotion detection

Pranaydeep Singh (UGent) , Orphée De Clercq (UGent) and Els Lefever (UGent)
Author
Organization
Abstract
The rapid advancement of multilingual pre-trained transformers has fueled significant progress in natural language understanding across diverse languages. Yet, their inner workings remain opaque, especially with regard to how individual neurons encode and generalize semantic and affective features across languages. This paper presents an interpretability study of a fine-tuned XLM-R model for multilingual emotion classification. Using neuron-level activation analysis, we investigate the variance of neurons across labels, cross-lingual alignment of activations, and the existence of “polyglot” versus language-specific neurons. Our results reveal that while certain neurons consistently encode emotion-related concepts across languages, others show strong monolingual specialization.
Keywords
LT3

Downloads

  • (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 453.97 KB

Citation

Please use this url to cite or link to this publication:

MLA
Singh, Pranaydeep, et al. “Lost in Activations : A Neuron-Level Analysis of Encoders for Cross-Lingual Emotion Detection.” Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers), Association for Computational Linguistics (ACL), 2026, pp. 154–59, doi:10.18653/v1/2026.eacl-short.9.
APA
Singh, P., De Clercq, O., & Lefever, E. (2026). Lost in activations : a neuron-level analysis of encoders for cross-lingual emotion detection. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers), 154–159. https://doi.org/10.18653/v1/2026.eacl-short.9
Chicago author-date
Singh, Pranaydeep, Orphée De Clercq, and Els Lefever. 2026. “Lost in Activations : A Neuron-Level Analysis of Encoders for Cross-Lingual Emotion Detection.” In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers), 154–59. Rabat, Morocco: Association for Computational Linguistics (ACL). https://doi.org/10.18653/v1/2026.eacl-short.9.
Chicago author-date (all authors)
Singh, Pranaydeep, Orphée De Clercq, and Els Lefever. 2026. “Lost in Activations : A Neuron-Level Analysis of Encoders for Cross-Lingual Emotion Detection.” In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers), 154–159. Rabat, Morocco: Association for Computational Linguistics (ACL). doi:10.18653/v1/2026.eacl-short.9.
Vancouver
1.
Singh P, De Clercq O, Lefever E. Lost in activations : a neuron-level analysis of encoders for cross-lingual emotion detection. In: Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (volume 2: short papers). Rabat, Morocco: Association for Computational Linguistics (ACL); 2026. p. 154–9.
IEEE
[1]
P. Singh, O. De Clercq, and E. Lefever, “Lost in activations : a neuron-level analysis of encoders for cross-lingual emotion detection,” in Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (volume 2: short papers), Rabat, Morocco, 2026, pp. 154–159.
@inproceedings{01KQD7AY59ACC912GHCE21P2T0,
  abstract     = {{The rapid advancement of multilingual pre-trained transformers has fueled significant progress in natural language understanding across diverse languages. Yet, their inner workings remain opaque, especially with regard to how individual neurons encode and generalize semantic and affective features across languages. This paper presents an interpretability study of a fine-tuned XLM-R model for multilingual emotion classification. Using neuron-level activation analysis, we investigate the variance of neurons across labels, cross-lingual alignment of activations, and the existence of “polyglot” versus language-specific neurons. Our results reveal that while certain neurons consistently encode emotion-related concepts across languages, others show strong monolingual specialization.}},
  author       = {{Singh, Pranaydeep and De Clercq, Orphée and Lefever, Els}},
  booktitle    = {{Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (volume 2: short papers)}},
  isbn         = {{9798891763814}},
  keywords     = {{LT3}},
  language     = {{eng}},
  location     = {{Rabat, Morocco}},
  pages        = {{154--159}},
  publisher    = {{Association for Computational Linguistics (ACL)}},
  title        = {{Lost in activations : a neuron-level analysis of encoders for cross-lingual emotion detection}},
  url          = {{http://doi.org/10.18653/v1/2026.eacl-short.9}},
  year         = {{2026}},
}

Altmetric
View in Altmetric