@inproceedings{icard-etal-2026-measuring,
    title = "Measuring Embedding Sensitivity to Authorial Style in {F}rench: Comparing Literary Texts with Language Model Rewritings",
    author = "Icard, Benjamin  and
      Sainero, Lila  and
      Breton, Alice  and
      Zve, Evangelia  and
      Ganascia, Jean-Gabriel",
    editor = {Hamilton, Sil  and
      {\"O}hman, Emily  and
      Hicke, Rebecca M. M.  and
      Bizzoni, Yuri  and
      Bax, Axel  and
      Matthews, Jacob A.  and
      H{\"a}m{\"a}l{\"a}inen, Mika},
    booktitle = "Proceedings of the 6th International Conference on Natural Language Processing for the Digital Humanities",
    month = jul,
    year = "2026",
    address = "San Diego, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.nlp4dh-1.8/",
    pages = "69--82",
    ISBN = "979-8-89176-427-9",
    abstract = "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."
}