@inproceedings{chen-etal-2026-old,
title = "Old Memories Die Hard: Unveiling Shallow Forgetting and Ripple Effects in Privacy Unlearning",
author = "Chen, Xiaoyi and
Wang, Haoyuan and
Tang, Siyuan and
Liu, Sijia and
Su, Liya and
Wang, XiaoFeng and
Tang, Haixu",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1476/",
pages = "31422--31440",
ISBN = "979-8-89176-462-0",
abstract = "Large language models (LLMs) often memorize personally identifiable information (PII) during training, raising serious privacy concerns. While machine unlearning has emerged as a promising mitigation, its effectiveness against privacy attacks remains unclear. We introduce PrivUn, an evaluation framework for privacy unlearning under recovery attacks, paired with three measurement tools: forgetting scores, association metrics, and layer-wise analysis of forgetting depth. This framework studies not only whether a model forgets, but also how forgetting propagates and where it fails.\par Using PrivUn, we make two findings. (i) The PII ripple effect is gradient-driven, not semantic. Graph-based proximity is uncorrelated with PII forgetting, while gradient-based association is more strongly correlated. (ii) Label-based unlearning is typically shallow, while PII is stored across deeper layers. Layer-wise evidence suggests that label-based methods mainly change the final layers, explaining why they reduce direct retrieval but remain vulnerable to stronger recovery attacks."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chen-etal-2026-old">
<titleInfo>
<title>Old Memories Die Hard: Unveiling Shallow Forgetting and Ripple Effects in Privacy Unlearning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xiaoyi</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haoyuan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Siyuan</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sijia</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liya</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">XiaoFeng</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haixu</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Large language models (LLMs) often memorize personally identifiable information (PII) during training, raising serious privacy concerns. While machine unlearning has emerged as a promising mitigation, its effectiveness against privacy attacks remains unclear. We introduce PrivUn, an evaluation framework for privacy unlearning under recovery attacks, paired with three measurement tools: forgetting scores, association metrics, and layer-wise analysis of forgetting depth. This framework studies not only whether a model forgets, but also how forgetting propagates and where it fails.\par Using PrivUn, we make two findings. (i) The PII ripple effect is gradient-driven, not semantic. Graph-based proximity is uncorrelated with PII forgetting, while gradient-based association is more strongly correlated. (ii) Label-based unlearning is typically shallow, while PII is stored across deeper layers. Layer-wise evidence suggests that label-based methods mainly change the final layers, explaining why they reduce direct retrieval but remain vulnerable to stronger recovery attacks.</abstract>
<identifier type="citekey">chen-etal-2026-old</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1476/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>31422</start>
<end>31440</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Old Memories Die Hard: Unveiling Shallow Forgetting and Ripple Effects in Privacy Unlearning
%A Chen, Xiaoyi
%A Wang, Haoyuan
%A Tang, Siyuan
%A Liu, Sijia
%A Su, Liya
%A Wang, XiaoFeng
%A Tang, Haixu
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F chen-etal-2026-old
%X Large language models (LLMs) often memorize personally identifiable information (PII) during training, raising serious privacy concerns. While machine unlearning has emerged as a promising mitigation, its effectiveness against privacy attacks remains unclear. We introduce PrivUn, an evaluation framework for privacy unlearning under recovery attacks, paired with three measurement tools: forgetting scores, association metrics, and layer-wise analysis of forgetting depth. This framework studies not only whether a model forgets, but also how forgetting propagates and where it fails.\par Using PrivUn, we make two findings. (i) The PII ripple effect is gradient-driven, not semantic. Graph-based proximity is uncorrelated with PII forgetting, while gradient-based association is more strongly correlated. (ii) Label-based unlearning is typically shallow, while PII is stored across deeper layers. Layer-wise evidence suggests that label-based methods mainly change the final layers, explaining why they reduce direct retrieval but remain vulnerable to stronger recovery attacks.
%U https://aclanthology.org/2026.findings-emnlp.1476/
%P 31422-31440
Markdown (Informal)
[Old Memories Die Hard: Unveiling Shallow Forgetting and Ripple Effects in Privacy Unlearning](https://aclanthology.org/2026.findings-emnlp.1476/) (Chen et al., Findings 2026)
ACL