StyleAdaptedLM: Enhancing Instruction Following Models with Efficient Stylistic Transfer
Authors:
- Pritika RamuPRPritika Ramu
- Apoorv SaxenaASApoorv Saxena
- Meghanath M YMYMeghanath M Y
- Varsha SankarVSVarsha Sankar
- Debraj BasuDBDebraj Basu
Preprint
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.Keywords
All Related Versions
This publication has 9 references indexed in Scilit:
- Lost in the Middle: How Language Models Use Long ContextsTransactions of the Association for Computational Linguistics, 2024
- TinyStyler: Efficient Few-Shot Text Style Transfer with Authorship Embeddings2024
- Distilling Text Style Transfer With Self-Explanation From LLMs2024
- Self-Instruct: Aligning Language Models with Self-Generated Instructions2023
- A Recipe for Arbitrary Text Style Transfer with Large Language Models2022
- Learning Universal Authorship Representations2021
- Prefix-Tuning: Optimizing Continuous Prompts for Generation2021
- A Deep Metric Learning Approach to Account Linking2021
- Dear Sir or Madam, May I Introduce the GYAFC Dataset: Corpus, Benchmarks and Metrics for Formality Style Transfer2018