@inproceedings{hong-etal-2024-cyclealign,
title = "{C}ycle{A}lign: Iterative Distillation from Black-box {LLM} to White-box Models for Better Human Alignment",
author = "Hong, Jixiang and
Tu, Quan and
Chen, Changyu and
Xing, Gao and
Zhang, Ji and
Yan, Rui",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.869",
doi = "10.18653/v1/2024.findings-acl.869",
pages = "14596--14609",
abstract = "Language models trained on large-scale corpus often generate harmful responses that are harmful and contrary to human values. A prevalent approach for human alignment is reinforcement learning from human feedback (RLHF), utilizing algorithms such as proximal policy optimization (PPO). However, these methods are often characterized by complexity, instability, and substantial resource consumption. Considering that existing large language models (LLMs) like ChatGPT are already relatively well-aligned and cost-friendly, researchers propose to align the language model with human preferences from AI feedback. Nevertheless, the common practices, that unidirectionally distill the responses, are constrained by the inherent capability of LLMs. To address it, we introduce CycleAlign, a framework that distills alignment capabilities from the parameter-invisible LLMs (black-box) to the parameter-visible models (white-box) in an iterative manner. CycleAlign iteratively improves both the white-box and black-box models by integrating static and dynamic in-context learning and a belief alignment method.Empirical results illustrate that the model fine-tuned by CycleAlign remarkably exceeds existing methods, and achieves the state-of-the-art performance in alignment with human value.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="hong-etal-2024-cyclealign">
<titleInfo>
<title>CycleAlign: Iterative Distillation from Black-box LLM to White-box Models for Better Human Alignment</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jixiang</namePart>
<namePart type="family">Hong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Quan</namePart>
<namePart type="family">Tu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Changyu</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gao</namePart>
<namePart type="family">Xing</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ji</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rui</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lun-Wei</namePart>
<namePart type="family">Ku</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Martins</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Srikumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Language models trained on large-scale corpus often generate harmful responses that are harmful and contrary to human values. A prevalent approach for human alignment is reinforcement learning from human feedback (RLHF), utilizing algorithms such as proximal policy optimization (PPO). However, these methods are often characterized by complexity, instability, and substantial resource consumption. Considering that existing large language models (LLMs) like ChatGPT are already relatively well-aligned and cost-friendly, researchers propose to align the language model with human preferences from AI feedback. Nevertheless, the common practices, that unidirectionally distill the responses, are constrained by the inherent capability of LLMs. To address it, we introduce CycleAlign, a framework that distills alignment capabilities from the parameter-invisible LLMs (black-box) to the parameter-visible models (white-box) in an iterative manner. CycleAlign iteratively improves both the white-box and black-box models by integrating static and dynamic in-context learning and a belief alignment method.Empirical results illustrate that the model fine-tuned by CycleAlign remarkably exceeds existing methods, and achieves the state-of-the-art performance in alignment with human value.</abstract>
<identifier type="citekey">hong-etal-2024-cyclealign</identifier>
<identifier type="doi">10.18653/v1/2024.findings-acl.869</identifier>
<location>
<url>https://aclanthology.org/2024.findings-acl.869</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>14596</start>
<end>14609</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CycleAlign: Iterative Distillation from Black-box LLM to White-box Models for Better Human Alignment
%A Hong, Jixiang
%A Tu, Quan
%A Chen, Changyu
%A Xing, Gao
%A Zhang, Ji
%A Yan, Rui
%Y Ku, Lun-Wei
%Y Martins, Andre
%Y Srikumar, Vivek
%S Findings of the Association for Computational Linguistics: ACL 2024
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F hong-etal-2024-cyclealign
%X Language models trained on large-scale corpus often generate harmful responses that are harmful and contrary to human values. A prevalent approach for human alignment is reinforcement learning from human feedback (RLHF), utilizing algorithms such as proximal policy optimization (PPO). However, these methods are often characterized by complexity, instability, and substantial resource consumption. Considering that existing large language models (LLMs) like ChatGPT are already relatively well-aligned and cost-friendly, researchers propose to align the language model with human preferences from AI feedback. Nevertheless, the common practices, that unidirectionally distill the responses, are constrained by the inherent capability of LLMs. To address it, we introduce CycleAlign, a framework that distills alignment capabilities from the parameter-invisible LLMs (black-box) to the parameter-visible models (white-box) in an iterative manner. CycleAlign iteratively improves both the white-box and black-box models by integrating static and dynamic in-context learning and a belief alignment method.Empirical results illustrate that the model fine-tuned by CycleAlign remarkably exceeds existing methods, and achieves the state-of-the-art performance in alignment with human value.
%R 10.18653/v1/2024.findings-acl.869
%U https://aclanthology.org/2024.findings-acl.869
%U https://doi.org/10.18653/v1/2024.findings-acl.869
%P 14596-14609
Markdown (Informal)
[CycleAlign: Iterative Distillation from Black-box LLM to White-box Models for Better Human Alignment](https://aclanthology.org/2024.findings-acl.869) (Hong et al., Findings 2024)
ACL