Style-Specific Neurons for Steering LLMs in Text Style Transfer
- URL: http://arxiv.org/abs/2410.00593v1
- Date: Tue, 1 Oct 2024 11:25:36 GMT
- Title: Style-Specific Neurons for Steering LLMs in Text Style Transfer
- Authors: Wen Lai, Viktor Hangya, Alexander Fraser,
- Abstract summary: Text style transfer (TST) aims to modify the style of a text without altering its original meaning.
We present sNeuron-TST, a novel approach for steering large language models using style-specific neurons.
- Score: 55.06697862691798
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: Text style transfer (TST) aims to modify the style of a text without altering its original meaning. Large language models (LLMs) demonstrate superior performance across multiple tasks, including TST. However, in zero-shot setups, they tend to directly copy a significant portion of the input text to the output without effectively changing its style. To enhance the stylistic variety and fluency of the text, we present sNeuron-TST, a novel approach for steering LLMs using style-specific neurons in TST. Specifically, we identify neurons associated with the source and target styles and deactivate source-style-only neurons to give target-style words a higher probability, aiming to enhance the stylistic diversity of the generated text. However, we find that this deactivation negatively impacts the fluency of the generated text, which we address by proposing an improved contrastive decoding method that accounts for rapid token probability shifts across layers caused by deactivated source-style neurons. Empirical experiments demonstrate the effectiveness of the proposed method on six benchmarks, encompassing formality, toxicity, politics, politeness, authorship, and sentiment.
Related papers
- Style Vectors for Steering Generative Large Language Model [1.4815455267254962]
We show that style vectors can be computed from recorded layer activations for input texts in a specific style.
The presented research constitutes a significant step towards developing more adaptive and effective AI-empowered interactive systems.
arXiv Detail & Related papers (2024-02-02T18:31:15Z) - Successor Features for Efficient Multisubject Controlled Text Generation [48.37713738712319]
We introduce SF-GEN, which is grounded in two primary concepts: successor features (SFs) and language model rectification.
SF-GEN seamlessly integrates the two to enable dynamic steering of text generation with no need to alter the LLM's parameters.
To the best of our knowledge, our research represents the first application of successor features in text generation.
arXiv Detail & Related papers (2023-11-03T00:17:08Z) - MSSRNet: Manipulating Sequential Style Representation for Unsupervised
Text Style Transfer [82.37710853235535]
Unsupervised text style transfer task aims to rewrite a text into target style while preserving its main content.
Traditional methods rely on the use of a fixed-sized vector to regulate text style, which is difficult to accurately convey the style strength for each individual token.
Our proposed method addresses this issue by assigning individual style vector to each token in a text, allowing for fine-grained control and manipulation of the style strength.
arXiv Detail & Related papers (2023-06-12T13:12:29Z) - VAE based Text Style Transfer with Pivot Words Enhancement Learning [5.717913255287939]
We propose a VAE based Text Style Transfer with pivOt Words Enhancement leaRning (VT-STOWER) method.
We introduce pivot words learning, which is applied to learn decisive words for a specific style.
The proposed VT-STOWER can be scaled to different TST scenarios with a novel and flexible style strength control mechanism.
arXiv Detail & Related papers (2021-12-06T16:41:26Z) - Fine-grained style control in Transformer-based Text-to-speech Synthesis [78.92428622630861]
We present a novel architecture to realize fine-grained style control on the Transformer-based text-to-speech synthesis (TransformerTTS)
We model the speaking style by extracting a time sequence of local style tokens (LST) from the reference speech.
Experiments show that with fine-grained style control, our system performs better in terms of naturalness, intelligibility, and style transferability.
arXiv Detail & Related papers (2021-10-12T19:50:02Z) - Neural Text Generation with Part-of-Speech Guided Softmax [82.63394952538292]
We propose using linguistic annotation, i.e., part-of-speech (POS), to guide the text generation.
We show that our proposed methods can generate more diverse text while maintaining comparable quality.
arXiv Detail & Related papers (2021-05-08T08:53:16Z) - StylePTB: A Compositional Benchmark for Fine-grained Controllable Text
Style Transfer [90.6768813620898]
Style transfer aims to controllably generate text with targeted stylistic changes while maintaining core meaning from the source sentence constant.
We introduce a large-scale benchmark, StylePTB, with paired sentences undergoing 21 fine-grained stylistic changes spanning atomic lexical, syntactic, semantic, and thematic transfers of text.
We find that existing methods on StylePTB struggle to model fine-grained changes and have an even more difficult time composing multiple styles.
arXiv Detail & Related papers (2021-04-12T04:25:09Z) - Syntax-driven Iterative Expansion Language Models for Controllable Text
Generation [2.578242050187029]
We propose a new paradigm for introducing a syntactic inductive bias into neural text generation.
Our experiments show that this paradigm is effective at text generation, with quality between LSTMs and Transformers, and comparable diversity.
arXiv Detail & Related papers (2020-04-05T14:29:40Z)
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.