@inproceedings{konen-etal-2024-style,
title = "Style Vectors for Steering Generative Large Language Models",
author = {Konen, Kai and
Jentzsch, Sophie and
Diallo, Diaoul{\'e} and
Sch{\"u}tt, Peer and
Bensch, Oliver and
El Baff, Roxanne and
Opitz, Dominik and
Hecking, Tobias},
editor = "Graham, Yvette and
Purver, Matthew",
booktitle = "Findings of the Association for Computational Linguistics: EACL 2024",
month = mar,
year = "2024",
address = "St. Julian{'}s, Malta",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-eacl.52",
pages = "782--802",
abstract = "This research explores strategies for steering the output of large language models (LLMs) towards specific styles, such as sentiment, emotion, or writing style, by adding style vectors to the activations of hidden layers during text generation. We show that style vectors can be simply computed from recorded layer activations for input texts in a specific style in contrast to more complex training-based approaches. Through a series of experiments, we demonstrate the effectiveness of activation engineering using such style vectors to influence the style of generated text in a nuanced and parameterisable way, distinguishing it from prompt engineering. The presented research constitutes a significant step towards developing more adaptive and effective AI-empowered interactive systems.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="konen-etal-2024-style">
<titleInfo>
<title>Style Vectors for Steering Generative Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">Konen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sophie</namePart>
<namePart type="family">Jentzsch</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Diaoulé</namePart>
<namePart type="family">Diallo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peer</namePart>
<namePart type="family">Schütt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oliver</namePart>
<namePart type="family">Bensch</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roxanne</namePart>
<namePart type="family">El Baff</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dominik</namePart>
<namePart type="family">Opitz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tobias</namePart>
<namePart type="family">Hecking</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-03</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EACL 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yvette</namePart>
<namePart type="family">Graham</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Matthew</namePart>
<namePart type="family">Purver</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">St. Julian’s, Malta</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>This research explores strategies for steering the output of large language models (LLMs) towards specific styles, such as sentiment, emotion, or writing style, by adding style vectors to the activations of hidden layers during text generation. We show that style vectors can be simply computed from recorded layer activations for input texts in a specific style in contrast to more complex training-based approaches. Through a series of experiments, we demonstrate the effectiveness of activation engineering using such style vectors to influence the style of generated text in a nuanced and parameterisable way, distinguishing it from prompt engineering. The presented research constitutes a significant step towards developing more adaptive and effective AI-empowered interactive systems.</abstract>
<identifier type="citekey">konen-etal-2024-style</identifier>
<location>
<url>https://aclanthology.org/2024.findings-eacl.52</url>
</location>
<part>
<date>2024-03</date>
<extent unit="page">
<start>782</start>
<end>802</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Style Vectors for Steering Generative Large Language Models
%A Konen, Kai
%A Jentzsch, Sophie
%A Diallo, Diaoulé
%A Schütt, Peer
%A Bensch, Oliver
%A El Baff, Roxanne
%A Opitz, Dominik
%A Hecking, Tobias
%Y Graham, Yvette
%Y Purver, Matthew
%S Findings of the Association for Computational Linguistics: EACL 2024
%D 2024
%8 March
%I Association for Computational Linguistics
%C St. Julian’s, Malta
%F konen-etal-2024-style
%X This research explores strategies for steering the output of large language models (LLMs) towards specific styles, such as sentiment, emotion, or writing style, by adding style vectors to the activations of hidden layers during text generation. We show that style vectors can be simply computed from recorded layer activations for input texts in a specific style in contrast to more complex training-based approaches. Through a series of experiments, we demonstrate the effectiveness of activation engineering using such style vectors to influence the style of generated text in a nuanced and parameterisable way, distinguishing it from prompt engineering. The presented research constitutes a significant step towards developing more adaptive and effective AI-empowered interactive systems.
%U https://aclanthology.org/2024.findings-eacl.52
%P 782-802
Markdown (Informal)
[Style Vectors for Steering Generative Large Language Models](https://aclanthology.org/2024.findings-eacl.52) (Konen et al., Findings 2024)
ACL
- Kai Konen, Sophie Jentzsch, Diaoulé Diallo, Peer Schütt, Oliver Bensch, Roxanne El Baff, Dominik Opitz, and Tobias Hecking. 2024. Style Vectors for Steering Generative Large Language Models. In Findings of the Association for Computational Linguistics: EACL 2024, pages 782–802, St. Julian’s, Malta. Association for Computational Linguistics.