Communication Dans Un Congrès Année : 2026

Measuring Embedding Sensitivity to Authorial Style in French: Comparing Literary Texts with Language Model Rewritings

Résumé

Large language models (LLMs) can convincingly imitate human writing styles, yet it remains unclear how much stylistic information is encoded in embeddings from any language model and retained after LLM rewriting. We investigate these questions in French, using a controlled literary dataset to quantify the effect of stylistic variation via changes in embedding dispersion. We observe that embeddings reliably capture authorial stylistic features and that these signals persist after rewriting, while also exhibiting LLM-specific patterns. These analytical results offer promising directions for authorship imitation detection in the era of language models.

Fichier principal
Vignette du fichier
NLP4DH_2026-51.pdf (664.99 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
licence

Dates et versions

hal-05618944 , version 1 (11-05-2026)

Licence

Identifiants

  • HAL Id : hal-05618944 , version 1

Citer

Benjamin Icard, Lila Sainero, Alice Breton, Evangelia Zve, Jean-Gabriel Ganascia. Measuring Embedding Sensitivity to Authorial Style in French: Comparing Literary Texts with Language Model Rewritings. 6th International Conference on Natural Language Processing for the Digital Humanities (NLP4DH 2026), Jul 2026, San Diego (California), United States. ⟨hal-05618944⟩
0 Consultations
0 Téléchargements

Partager

  • More