@inproceedings{wan-etal-2025-every,
title = "Not Every Token Needs Forgetting: Selective Unlearning Balancing Forgetting and Utility in Large Language Models",
author = "Wan, Yixin and
Ramakrishna, Anil and
Chang, Kai-Wei and
Cevher, Volkan and
Gupta, Rahul",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-emnlp.96/",
pages = "1827--1835",
ISBN = "979-8-89176-335-7",
abstract = "Large Language Model (LLM) unlearning has recently gained significant attention, driven by the need to remove unwanted information{---}such as private, sensitive, or copyrighted content{---}from trained models. However, conventional unlearning approaches indiscriminately update model parameters to forget all tokens in a target document, including common tokens (e.g., pronouns, prepositions, general nouns) that carry general knowledge. In this paper, we highlight that ``not every token needs forgetting''. We propose **Selective Unlearning (SU)**, which identifies a critical subset of tokens within the forgetting set that is relevant to the unwanted information, and unlearns only those tokens. Experiments on two benchmarks and six baseline unlearning algorithms demonstrate that SU not only achieves effective unlearning on the targeted forget data, but also significantly preserves the model{'}s utility in the retaining set."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wan-etal-2025-every">
<titleInfo>
<title>Not Every Token Needs Forgetting: Selective Unlearning Balancing Forgetting and Utility in Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yixin</namePart>
<namePart type="family">Wan</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">Kai-Wei</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Volkan</namePart>
<namePart type="family">Cevher</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>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christos</namePart>
<namePart type="family">Christodoulopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolyn</namePart>
<namePart type="family">Rose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Violet</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-335-7</identifier>
</relatedItem>
<abstract>Large Language Model (LLM) unlearning has recently gained significant attention, driven by the need to remove unwanted information—such as private, sensitive, or copyrighted content—from trained models. However, conventional unlearning approaches indiscriminately update model parameters to forget all tokens in a target document, including common tokens (e.g., pronouns, prepositions, general nouns) that carry general knowledge. In this paper, we highlight that “not every token needs forgetting”. We propose **Selective Unlearning (SU)**, which identifies a critical subset of tokens within the forgetting set that is relevant to the unwanted information, and unlearns only those tokens. Experiments on two benchmarks and six baseline unlearning algorithms demonstrate that SU not only achieves effective unlearning on the targeted forget data, but also significantly preserves the model’s utility in the retaining set.</abstract>
<identifier type="citekey">wan-etal-2025-every</identifier>
<location>
<url>https://aclanthology.org/2025.findings-emnlp.96/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>1827</start>
<end>1835</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Not Every Token Needs Forgetting: Selective Unlearning Balancing Forgetting and Utility in Large Language Models
%A Wan, Yixin
%A Ramakrishna, Anil
%A Chang, Kai-Wei
%A Cevher, Volkan
%A Gupta, Rahul
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Findings of the Association for Computational Linguistics: EMNLP 2025
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-335-7
%F wan-etal-2025-every
%X Large Language Model (LLM) unlearning has recently gained significant attention, driven by the need to remove unwanted information—such as private, sensitive, or copyrighted content—from trained models. However, conventional unlearning approaches indiscriminately update model parameters to forget all tokens in a target document, including common tokens (e.g., pronouns, prepositions, general nouns) that carry general knowledge. In this paper, we highlight that “not every token needs forgetting”. We propose **Selective Unlearning (SU)**, which identifies a critical subset of tokens within the forgetting set that is relevant to the unwanted information, and unlearns only those tokens. Experiments on two benchmarks and six baseline unlearning algorithms demonstrate that SU not only achieves effective unlearning on the targeted forget data, but also significantly preserves the model’s utility in the retaining set.
%U https://aclanthology.org/2025.findings-emnlp.96/
%P 1827-1835
Markdown (Informal)
[Not Every Token Needs Forgetting: Selective Unlearning Balancing Forgetting and Utility in Large Language Models](https://aclanthology.org/2025.findings-emnlp.96/) (Wan et al., Findings 2025)
ACL