@inproceedings{kim-etal-2026-confess,
title = "Confess What You Know: Forget-Set Misalignment with Model Knowledge in {LLM} Unlearning",
author = "Kim, Miso and
Lee, Georu and
Jeong, Seungwon and
Lee, Woojin",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1839/",
pages = "41825--41846",
ISBN = "979-8-89176-459-0",
abstract = "Machine unlearning for large language models (LLMs) often assumes that a pre-defined forget set matches what the model has memorized, but this frequently breaks in realistic privacy settings where the original training data is inaccessible. We term this gap forget-set misalignment and identify two cases. In *Under Unlearning*, the forget set omits memorized information and leakage persists. In *Out-of-Knowledge Unlearning*, the algorithm is driven to ``forget'' knowledge the model never learned, perturbing parameters and degrading utility. Using gradient-level analysis, we show these behaviors arise from misaligned unlearning targets rather than specific optimization choices. We then propose **CONfession-to-Forget-Set (CONFS)**, a *data-blind* framework that constructs model-aligned forget sets by eliciting and formalizing the model{'}s memorized knowledge. Across synthetic, multimodal, and real-world benchmarks, CONFS approaches Gold-standard performance on several metrics and achieves a competitive forgetting-utility balance, while preserving utility better than other data-blind forget-set constructions."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kim-etal-2026-confess">
<titleInfo>
<title>Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Miso</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Georu</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seungwon</namePart>
<namePart type="family">Jeong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Woojin</namePart>
<namePart type="family">Lee</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Machine unlearning for large language models (LLMs) often assumes that a pre-defined forget set matches what the model has memorized, but this frequently breaks in realistic privacy settings where the original training data is inaccessible. We term this gap forget-set misalignment and identify two cases. In *Under Unlearning*, the forget set omits memorized information and leakage persists. In *Out-of-Knowledge Unlearning*, the algorithm is driven to “forget” knowledge the model never learned, perturbing parameters and degrading utility. Using gradient-level analysis, we show these behaviors arise from misaligned unlearning targets rather than specific optimization choices. We then propose **CONfession-to-Forget-Set (CONFS)**, a *data-blind* framework that constructs model-aligned forget sets by eliciting and formalizing the model’s memorized knowledge. Across synthetic, multimodal, and real-world benchmarks, CONFS approaches Gold-standard performance on several metrics and achieves a competitive forgetting-utility balance, while preserving utility better than other data-blind forget-set constructions.</abstract>
<identifier type="citekey">kim-etal-2026-confess</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1839/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>41825</start>
<end>41846</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearning
%A Kim, Miso
%A Lee, Georu
%A Jeong, Seungwon
%A Lee, Woojin
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F kim-etal-2026-confess
%X Machine unlearning for large language models (LLMs) often assumes that a pre-defined forget set matches what the model has memorized, but this frequently breaks in realistic privacy settings where the original training data is inaccessible. We term this gap forget-set misalignment and identify two cases. In *Under Unlearning*, the forget set omits memorized information and leakage persists. In *Out-of-Knowledge Unlearning*, the algorithm is driven to “forget” knowledge the model never learned, perturbing parameters and degrading utility. Using gradient-level analysis, we show these behaviors arise from misaligned unlearning targets rather than specific optimization choices. We then propose **CONfession-to-Forget-Set (CONFS)**, a *data-blind* framework that constructs model-aligned forget sets by eliciting and formalizing the model’s memorized knowledge. Across synthetic, multimodal, and real-world benchmarks, CONFS approaches Gold-standard performance on several metrics and achieves a competitive forgetting-utility balance, while preserving utility better than other data-blind forget-set constructions.
%U https://aclanthology.org/2026.emnlp-main.1839/
%P 41825-41846
Markdown (Informal)
[Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearning](https://aclanthology.org/2026.emnlp-main.1839/) (Kim et al., EMNLP 2026)
ACL