@inproceedings{vasilev-etal-2025-unilogit,
title = "Unilogit: Robust Machine Unlearning for {LLM}s Using Uniform-Target Self-Distillation",
author = "Vasilev, Stefan and
Herold, Christian and
Liao, Baohao and
Hashemi, Seyyed Hadi and
Khadivi, Shahram and
Monz, Christof",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-acl.1154/",
doi = "10.18653/v1/2025.findings-acl.1154",
pages = "22453--22472",
ISBN = "979-8-89176-256-5",
abstract = "This paper introduces Unilogit, a novel self-distillation method for machine unlearning in Large Language Models. Unilogit addresses the challenge of selectively forgetting specific information while maintaining overall model utility, a critical task in compliance with data privacy regulations like GDPR. Unlike prior methods that rely on static hyperparameters or starting model outputs, Unilogit dynamically adjusts target logits to achieve a uniform probability for the target token, leveraging the current model{'}s outputs for more accurate self-distillation targets. This approach not only eliminates the need for additional hyperparameters but also enhances the model{'}s ability to approximate the golden targets. Extensive experiments on public benchmarks and an in-house e-commerce dataset demonstrate Unilogit{'}s superior performance in balancing forget and retain objectives, outperforming state-of-the-art methods such as NPO and UnDIAL. Our analysis further reveals Unilogit{'}s robustness across various scenarios, highlighting its practical applicability and effectiveness in achieving efficacious machine unlearning."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vasilev-etal-2025-unilogit">
<titleInfo>
<title>Unilogit: Robust Machine Unlearning for LLMs Using Uniform-Target Self-Distillation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stefan</namePart>
<namePart type="family">Vasilev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christian</namePart>
<namePart type="family">Herold</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Baohao</namePart>
<namePart type="family">Liao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seyyed</namePart>
<namePart type="given">Hadi</namePart>
<namePart type="family">Hashemi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shahram</namePart>
<namePart type="family">Khadivi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christof</namePart>
<namePart type="family">Monz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-256-5</identifier>
</relatedItem>
<abstract>This paper introduces Unilogit, a novel self-distillation method for machine unlearning in Large Language Models. Unilogit addresses the challenge of selectively forgetting specific information while maintaining overall model utility, a critical task in compliance with data privacy regulations like GDPR. Unlike prior methods that rely on static hyperparameters or starting model outputs, Unilogit dynamically adjusts target logits to achieve a uniform probability for the target token, leveraging the current model’s outputs for more accurate self-distillation targets. This approach not only eliminates the need for additional hyperparameters but also enhances the model’s ability to approximate the golden targets. Extensive experiments on public benchmarks and an in-house e-commerce dataset demonstrate Unilogit’s superior performance in balancing forget and retain objectives, outperforming state-of-the-art methods such as NPO and UnDIAL. Our analysis further reveals Unilogit’s robustness across various scenarios, highlighting its practical applicability and effectiveness in achieving efficacious machine unlearning.</abstract>
<identifier type="citekey">vasilev-etal-2025-unilogit</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.1154</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.1154/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>22453</start>
<end>22472</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Unilogit: Robust Machine Unlearning for LLMs Using Uniform-Target Self-Distillation
%A Vasilev, Stefan
%A Herold, Christian
%A Liao, Baohao
%A Hashemi, Seyyed Hadi
%A Khadivi, Shahram
%A Monz, Christof
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F vasilev-etal-2025-unilogit
%X This paper introduces Unilogit, a novel self-distillation method for machine unlearning in Large Language Models. Unilogit addresses the challenge of selectively forgetting specific information while maintaining overall model utility, a critical task in compliance with data privacy regulations like GDPR. Unlike prior methods that rely on static hyperparameters or starting model outputs, Unilogit dynamically adjusts target logits to achieve a uniform probability for the target token, leveraging the current model’s outputs for more accurate self-distillation targets. This approach not only eliminates the need for additional hyperparameters but also enhances the model’s ability to approximate the golden targets. Extensive experiments on public benchmarks and an in-house e-commerce dataset demonstrate Unilogit’s superior performance in balancing forget and retain objectives, outperforming state-of-the-art methods such as NPO and UnDIAL. Our analysis further reveals Unilogit’s robustness across various scenarios, highlighting its practical applicability and effectiveness in achieving efficacious machine unlearning.
%R 10.18653/v1/2025.findings-acl.1154
%U https://aclanthology.org/2025.findings-acl.1154/
%U https://doi.org/10.18653/v1/2025.findings-acl.1154
%P 22453-22472
Markdown (Informal)
[Unilogit: Robust Machine Unlearning for LLMs Using Uniform-Target Self-Distillation](https://aclanthology.org/2025.findings-acl.1154/) (Vasilev et al., Findings 2025)
ACL