StyleAdaptedLM: Enhancing Instruction Following Models with Efficient Stylistic Transfer

2025-07-24

AI Summary

Large language models (LLMs) struggle to simultaneously achieve stylistic adaptation (e.g., brand tone) and faithful instruction following when instruction-response paired data is scarce. Method: We propose a two-stage LoRA-based style transfer framework: Stage I disentangles stylistic representations from unpaired, unstructured text; Stage II seamlessly integrates the learned style module into a pretrained instruction-tuned model—without requiring instruction-format annotations or compromising task capability. Contribution/Results: Our approach eliminates interference between style and task learning, enabling lightweight, controllable, and reusable style adaptation without paired corpora—the first such decoupled style-task adaptation method. Experiments across multiple open-source LLMs and benchmarks demonstrate significant improvements in style consistency (validated by both automated metrics and human evaluation), while fully preserving original instruction-following performance.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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.

Problem

• Adapting LLMs to specific stylistic traits without paired data
• Transferring stylistic characteristics without sacrificing instruction adherence
• Enhancing brand voice customization in enterprise communication

Innovation

• Uses Low-Rank Adaptation for style transfer
• Trains LoRA adapters on unstructured stylistic corpora
• Merges adapters with instruction-following models