StyleAdaptedLM: Enhancing Instruction Following Models with Efficient Stylistic Transfer

TL;DR

StyleAdaptedLM is a framework that efficiently personalizes large language models with specific stylistic traits using low-rank adaptation, maintaining instruction-following capabilities without requiring paired data.

Contribution

It introduces a novel method combining LoRA adapters with instruction models for effective stylistic transfer without data pairing or performance loss.

Findings

Improved stylistic consistency confirmed by human evaluations

Maintains instruction-following performance across datasets

Enables efficient stylistic customization without paired data

Abstract

Adapting LLMs to specific stylistic characteristics, like brand voice or authorial tones, is crucial for enterprise communication but challenging to achieve from corpora which lacks instruction-response formatting without compromising instruction adherence. We introduce StyleAdaptedLM, a framework that efficiently transfers stylistic traits to instruction-following models using Low-Rank Adaptation (LoRA). LoRA adapters are first trained on a base model with diverse unstructured stylistic corpora, then merged with a separate instruction-following model. This enables robust stylistic customization without paired data or sacrificing task performance. Experiments across multiple datasets and models demonstrate improved stylistic consistency while preserving instruction adherence, with human evaluations confirming brand-specific convention uptake. StyleAdaptedLM offers an efficient path for stylistic personalization in LLMs.