@inproceedings{kim-etal-2023-critic,
title = "Critic-Guided Decoding for Controlled Text Generation",
author = "Kim, Minbeom and
Lee, Hwanhee and
Yoo, Kang Min and
Park, Joonsuk and
Lee, Hwaran and
Jung, Kyomin",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-acl.281",
doi = "10.18653/v1/2023.findings-acl.281",
pages = "4598--4612",
abstract = "Steering language generation towards objectives or away from undesired content has been a long-standing goal in utilizing language models (LM). Recent work has demonstrated reinforcement learning and weighted decoding as effective approaches to achieve a higher level of language control and quality with pros and cons. In this work, we propose a novel critic decoding method for controlled language generation (CriticControl) that combines the strengths of reinforcement learning and weighted decoding. Specifically, we adopt the actor-critic framework and train an LM-steering critic from reward models. Similar to weighted decoding, our method freezes the language model and manipulates the output token distribution using a critic to improve training efficiency and stability. Evaluation of our method on three controlled generation tasks, topic control, sentiment control, and detoxification, shows that our approach generates more coherent and well-controlled texts than previous methods. In addition, CriticControl demonstrates superior generalization ability in zero-shot settings. Human evaluation studies also corroborate our findings.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kim-etal-2023-critic">
<titleInfo>
<title>Critic-Guided Decoding for Controlled Text Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Minbeom</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hwanhee</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kang</namePart>
<namePart type="given">Min</namePart>
<namePart type="family">Yoo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joonsuk</namePart>
<namePart type="family">Park</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hwaran</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kyomin</namePart>
<namePart type="family">Jung</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2023</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jordan</namePart>
<namePart type="family">Boyd-Graber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoaki</namePart>
<namePart type="family">Okazaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Steering language generation towards objectives or away from undesired content has been a long-standing goal in utilizing language models (LM). Recent work has demonstrated reinforcement learning and weighted decoding as effective approaches to achieve a higher level of language control and quality with pros and cons. In this work, we propose a novel critic decoding method for controlled language generation (CriticControl) that combines the strengths of reinforcement learning and weighted decoding. Specifically, we adopt the actor-critic framework and train an LM-steering critic from reward models. Similar to weighted decoding, our method freezes the language model and manipulates the output token distribution using a critic to improve training efficiency and stability. Evaluation of our method on three controlled generation tasks, topic control, sentiment control, and detoxification, shows that our approach generates more coherent and well-controlled texts than previous methods. In addition, CriticControl demonstrates superior generalization ability in zero-shot settings. Human evaluation studies also corroborate our findings.</abstract>
<identifier type="citekey">kim-etal-2023-critic</identifier>
<identifier type="doi">10.18653/v1/2023.findings-acl.281</identifier>
<location>
<url>https://aclanthology.org/2023.findings-acl.281</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>4598</start>
<end>4612</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Critic-Guided Decoding for Controlled Text Generation
%A Kim, Minbeom
%A Lee, Hwanhee
%A Yoo, Kang Min
%A Park, Joonsuk
%A Lee, Hwaran
%A Jung, Kyomin
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Findings of the Association for Computational Linguistics: ACL 2023
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F kim-etal-2023-critic
%X Steering language generation towards objectives or away from undesired content has been a long-standing goal in utilizing language models (LM). Recent work has demonstrated reinforcement learning and weighted decoding as effective approaches to achieve a higher level of language control and quality with pros and cons. In this work, we propose a novel critic decoding method for controlled language generation (CriticControl) that combines the strengths of reinforcement learning and weighted decoding. Specifically, we adopt the actor-critic framework and train an LM-steering critic from reward models. Similar to weighted decoding, our method freezes the language model and manipulates the output token distribution using a critic to improve training efficiency and stability. Evaluation of our method on three controlled generation tasks, topic control, sentiment control, and detoxification, shows that our approach generates more coherent and well-controlled texts than previous methods. In addition, CriticControl demonstrates superior generalization ability in zero-shot settings. Human evaluation studies also corroborate our findings.
%R 10.18653/v1/2023.findings-acl.281
%U https://aclanthology.org/2023.findings-acl.281
%U https://doi.org/10.18653/v1/2023.findings-acl.281
%P 4598-4612
Markdown (Informal)
[Critic-Guided Decoding for Controlled Text Generation](https://aclanthology.org/2023.findings-acl.281) (Kim et al., Findings 2023)
ACL
- Minbeom Kim, Hwanhee Lee, Kang Min Yoo, Joonsuk Park, Hwaran Lee, and Kyomin Jung. 2023. Critic-Guided Decoding for Controlled Text Generation. In Findings of the Association for Computational Linguistics: ACL 2023, pages 4598–4612, Toronto, Canada. Association for Computational Linguistics.