AuthorMix: Modular Authorship Style Transfer via Layer-wise Adapter Mixing

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Thillainathan, Sarubi, Lee, Ji-Ung, Sullivan, Michael, Koller, Alexander
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917543678050304
author Thillainathan, Sarubi
Lee, Ji-Ung
Sullivan, Michael
Koller, Alexander
author_facet Thillainathan, Sarubi
Lee, Ji-Ung
Sullivan, Michael
Koller, Alexander
contents The task of authorship style transfer involves rewriting text in the style of a target author while preserving the meaning of the original text. Existing style transfer methods train a single model on large corpora to model all target styles at once: this high-cost approach offers limited flexibility for target-specific adaptation, and often sacrifices meaning preservation for style transfer. In this paper, we propose AuthorMix: a lightweight, modular, and interpretable style transfer framework. We train individual, style-specific LoRA adapters on a small set of high-resource authors, allowing the rapid training of specialized adaptation models for each new target via learned, layer-wise adapter mixing, using only a handful of target-style training examples. AuthorMix outperforms existing, SoTA style-transfer baselines-as well as GPT-5.1-for low-resource targets, achieving the highest overall score and substantially improving meaning preservation in both automatic and human evaluations.
format Preprint
id arxiv_https___arxiv_org_abs_2603_23069
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle AuthorMix: Modular Authorship Style Transfer via Layer-wise Adapter Mixing
Thillainathan, Sarubi
Lee, Ji-Ung
Sullivan, Michael
Koller, Alexander
Computation and Language
Artificial Intelligence
The task of authorship style transfer involves rewriting text in the style of a target author while preserving the meaning of the original text. Existing style transfer methods train a single model on large corpora to model all target styles at once: this high-cost approach offers limited flexibility for target-specific adaptation, and often sacrifices meaning preservation for style transfer. In this paper, we propose AuthorMix: a lightweight, modular, and interpretable style transfer framework. We train individual, style-specific LoRA adapters on a small set of high-resource authors, allowing the rapid training of specialized adaptation models for each new target via learned, layer-wise adapter mixing, using only a handful of target-style training examples. AuthorMix outperforms existing, SoTA style-transfer baselines-as well as GPT-5.1-for low-resource targets, achieving the highest overall score and substantially improving meaning preservation in both automatic and human evaluations.
title AuthorMix: Modular Authorship Style Transfer via Layer-wise Adapter Mixing
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2603.23069