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

TL;DR

AuthorMix introduces a modular, layer-wise adapter mixing approach for authorship style transfer, enabling efficient, high-quality style adaptation with minimal data and outperforming existing methods.

Contribution

It proposes a novel, lightweight framework using style-specific adapters and layer-wise mixing for flexible, low-resource authorship style transfer.

Findings

Outperforms existing style transfer baselines and GPT-5.1 on low-resource targets.

Achieves highest overall score and better meaning preservation.

Enables rapid training of style-specific models with few examples.

Abstract

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.