@inproceedings{meng-etal-2024-attribute,
title = "Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification",
author = "Meng, Tao and
Mehrabi, Ninareh and
Goyal, Palash and
Ramakrishna, Anil and
Galstyan, Aram and
Zemel, Richard and
Chang, Kai-Wei and
Gupta, Rahul and
Peris, Charith",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-emnlp.779",
pages = "13329--13341",
abstract = "We propose a constraint learning schema forfine-tuning Large Language Models (LLMs)with attribute control. Given a training corpusand control criteria formulated as a sequence-level constraint on model outputs, our methodfine-tunes the LLM on the training corpus whileenhancing constraint satisfaction with minimalimpact on its utility and generation quality.Specifically, our approach regularizes the LLMtraining by penalizing the KL divergence be-tween the desired output distribution, which sat-isfies the constraints, and the LLM{'}s posterior.This regularization term can be approximatedby an auxiliary model trained to decomposethe sequence-level constraints into token-levelguidance, allowing the term to be measuredby a closed-form formulation. To further im-prove efficiency, we design a parallel schemefor concurrently updating both the LLM andthe auxiliary model. We evaluate the empiricalperformance of our approach by controlling thetoxicity when training an LLM. We show thatour approach leads to an LLM that producesfewer inappropriate responses while achievingcompetitive performance on benchmarks and atoxicity detection task",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="meng-etal-2024-attribute">
<titleInfo>
<title>Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tao</namePart>
<namePart type="family">Meng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ninareh</namePart>
<namePart type="family">Mehrabi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Palash</namePart>
<namePart type="family">Goyal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anil</namePart>
<namePart type="family">Ramakrishna</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aram</namePart>
<namePart type="family">Galstyan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Richard</namePart>
<namePart type="family">Zemel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai-Wei</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rahul</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Charith</namePart>
<namePart type="family">Peris</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yaser</namePart>
<namePart type="family">Al-Onaizan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun-Nung</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>We propose a constraint learning schema forfine-tuning Large Language Models (LLMs)with attribute control. Given a training corpusand control criteria formulated as a sequence-level constraint on model outputs, our methodfine-tunes the LLM on the training corpus whileenhancing constraint satisfaction with minimalimpact on its utility and generation quality.Specifically, our approach regularizes the LLMtraining by penalizing the KL divergence be-tween the desired output distribution, which sat-isfies the constraints, and the LLM’s posterior.This regularization term can be approximatedby an auxiliary model trained to decomposethe sequence-level constraints into token-levelguidance, allowing the term to be measuredby a closed-form formulation. To further im-prove efficiency, we design a parallel schemefor concurrently updating both the LLM andthe auxiliary model. We evaluate the empiricalperformance of our approach by controlling thetoxicity when training an LLM. We show thatour approach leads to an LLM that producesfewer inappropriate responses while achievingcompetitive performance on benchmarks and atoxicity detection task</abstract>
<identifier type="citekey">meng-etal-2024-attribute</identifier>
<location>
<url>https://aclanthology.org/2024.findings-emnlp.779</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>13329</start>
<end>13341</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification
%A Meng, Tao
%A Mehrabi, Ninareh
%A Goyal, Palash
%A Ramakrishna, Anil
%A Galstyan, Aram
%A Zemel, Richard
%A Chang, Kai-Wei
%A Gupta, Rahul
%A Peris, Charith
%Y Al-Onaizan, Yaser
%Y Bansal, Mohit
%Y Chen, Yun-Nung
%S Findings of the Association for Computational Linguistics: EMNLP 2024
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, USA
%F meng-etal-2024-attribute
%X We propose a constraint learning schema forfine-tuning Large Language Models (LLMs)with attribute control. Given a training corpusand control criteria formulated as a sequence-level constraint on model outputs, our methodfine-tunes the LLM on the training corpus whileenhancing constraint satisfaction with minimalimpact on its utility and generation quality.Specifically, our approach regularizes the LLMtraining by penalizing the KL divergence be-tween the desired output distribution, which sat-isfies the constraints, and the LLM’s posterior.This regularization term can be approximatedby an auxiliary model trained to decomposethe sequence-level constraints into token-levelguidance, allowing the term to be measuredby a closed-form formulation. To further im-prove efficiency, we design a parallel schemefor concurrently updating both the LLM andthe auxiliary model. We evaluate the empiricalperformance of our approach by controlling thetoxicity when training an LLM. We show thatour approach leads to an LLM that producesfewer inappropriate responses while achievingcompetitive performance on benchmarks and atoxicity detection task
%U https://aclanthology.org/2024.findings-emnlp.779
%P 13329-13341
Markdown (Informal)
[Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification](https://aclanthology.org/2024.findings-emnlp.779) (Meng et al., Findings 2024)
ACL
- Tao Meng, Ninareh Mehrabi, Palash Goyal, Anil Ramakrishna, Aram Galstyan, Richard Zemel, Kai-Wei Chang, Rahul Gupta, and Charith Peris. 2024. Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13329–13341, Miami, Florida, USA. Association for Computational Linguistics.