Interpreting Style Representations via Style-Eliciting Prompts

• URL: http://arxiv.org/abs/2606.05716v1
• Date: Thu, 04 Jun 2026 05:17:23 GMT
• Title: Interpreting Style Representations via Style-Eliciting Prompts
• Abstract summary: Style representation learning is a powerful tool for authorship analysis and modeling writing style.<n>Recent work has attempted to explain these representations by generating natural language descriptions with large language models conditioned on input text.<n>We propose a novel framework for interpreting style representations through style-eliciting prompts.
• Score: 33.42332250065166
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Style representation learning is a powerful tool for authorship analysis and modeling writing style, yet the latent nature of learned representations makes them difficult to interpret. Recent work has attempted to explain these representations by generating natural language descriptions with large language models (LLMs) conditioned on input text. However, such descriptions are often prone to the LLM's biases and hallucinations, and they lack an explicit objective and practical utility. In this work, we propose a novel framework for interpreting style representations through style-eliciting prompts: natural language instructions designed to steer LLMs to generate text that reflects specific stylistic attributes. We curate 1,010 distinct style features spanning 26 stylistic categories and construct a dataset by prompting an LLM to generate text conditioned on these features. Using this data, we train a decoder to generate a style prompt from the style representation of the generated text. We evaluate our approach on three tasks: (1) recovering original style prompts from generated text, (2) generating text in the same style using the recovered prompts, and (3) steering LLM outputs to match the style of human-written texts. Experiments demonstrate that our method consistently outperforms strong baselines that directly prompt LLMs with target text, achieving superior performance in both style description and style imitation. These results highlight that style-eliciting prompts can provide a practical and interpretable interface to stylistic information encoded in style representations.

Related papers

• Interpretable Stylistic Variation in Human and LLM Writing Across Genres, Models, and Decoding Strategies [0.34202935599316525]
  Large Language Models (LLMs) are now capable of generating highly fluent, human-like text.<n>LLMs enable many applications, but also raise concerns such as large scale spam, phishing, or academic misuse.<n>We perform a large scale analysis of stylistic variation across human-written text and outputs from 11 LLMs.
  arXiv  Detail & Related papers  (2026-04-15T17:31:21Z)
• Personalized Image Generation from an Author Writing Style [0.29998889086656577]
  Translating nuanced, textually-defined authorial writing styles into compelling visual representations presents a novel challenge in generative AI.<n>This paper introduces a pipeline that leverages Author Writing Sheets (AWS) as input to a Large Language Model (LLM)<n>We evaluated our approach using 49 author styles from Reddit data, with human evaluators assessing the stylistic match and visual distinctiveness of the generated images.
  arXiv  Detail & Related papers  (2025-07-04T05:53:48Z)
• DRESSing Up LLM: Efficient Stylized Question-Answering via Style Subspace Editing [23.467409274256255]
  DRESS is a novel approach for generating stylized large language model (LLM) responses through representation editing.<n>Our approach disentangles a style-relevant subspace within the model's representation space to conduct representation editing.<n>In short, DRESS is a lightweight, train-free solution for enhancing LLMs with flexible and effective style control.
  arXiv  Detail & Related papers  (2025-01-24T10:04:53Z)
• Learning to Generate Text in Arbitrary Writing Styles [6.7308816341849695]
  It is desirable for language models to produce text in an author-specific style on the basis of a potentially small writing sample. We propose to guide a language model to generate text in a target style using contrastively-trained representations that capture stylometric features.
  arXiv  Detail & Related papers  (2023-12-28T18:58:52Z)
• Visual Captioning at Will: Describing Images and Videos Guided by a Few Stylized Sentences [49.66987347397398]
  Few-Shot Stylized Visual Captioning aims to generate captions in any desired style, using only a few examples as guidance during inference. We propose a framework called FS-StyleCap for this task, which utilizes a conditional encoder-decoder language model and a visual projection module.
  arXiv  Detail & Related papers  (2023-07-31T04:26:01Z)
• CLIP4STR: A Simple Baseline for Scene Text Recognition with Pre-trained Vision-Language Model [55.321010757641524]
  We introduce CLIP4STR, a simple yet effective STR method built upon image and text encoders of CLIP.<n>We scale CLIP4STR in terms of the model size, pre-training data, and training data, achieving state-of-the-art performance on 13 STR benchmarks.
  arXiv  Detail & Related papers  (2023-05-23T12:51:20Z)
• Learning Interpretable Style Embeddings via Prompting LLMs [46.74488355350601]
  Style representation learning builds content-independent representations of author style in text. Current style representation learning uses neural methods to disentangle style from content to create style vectors. We use prompting to perform stylometry on a large number of texts to create a synthetic dataset and train human-interpretable style representations.
  arXiv  Detail & Related papers  (2023-05-22T04:07:54Z)
• Stylized Data-to-Text Generation: A Case Study in the E-Commerce Domain [53.22419717434372]
  We propose a new task, namely stylized data-to-text generation, whose aim is to generate coherent text according to a specific style. This task is non-trivial, due to three challenges: the logic of the generated text, unstructured style reference, and biased training samples. We propose a novel stylized data-to-text generation model, named StyleD2T, comprising three components: logic planning-enhanced data embedding, mask-based style embedding, and unbiased stylized text generation.
  arXiv  Detail & Related papers  (2023-05-05T03:02:41Z)
• StylerDALLE: Language-Guided Style Transfer Using a Vector-Quantized Tokenizer of a Large-Scale Generative Model [64.26721402514957]
  We propose StylerDALLE, a style transfer method that uses natural language to describe abstract art styles. Specifically, we formulate the language-guided style transfer task as a non-autoregressive token sequence translation. To incorporate style information, we propose a Reinforcement Learning strategy with CLIP-based language supervision.
  arXiv  Detail & Related papers  (2023-03-16T12:44:44Z)
• StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery [71.1862388442953]
  We develop a text-based interface for StyleGAN image manipulation. We first introduce an optimization scheme that utilizes a CLIP-based loss to modify an input latent vector in response to a user-provided text prompt. Next, we describe a latent mapper that infers a text-guided latent manipulation step for a given input image, allowing faster and more stable text-based manipulation.
  arXiv  Detail & Related papers  (2021-03-31T17:51:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.