@inproceedings{nguyen-etal-2026-flour,
title = "{FLOUR}: Mixing the Right Ingredients for {LLM} Unlearning in Latent Space",
author = "Nguyen, Thuy Dung and
Nguyen, Giang and
Johnson, Taylor T and
Leach, Kevin",
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.2376/",
doi = "10.18653/2026.findings-emnlp.2376",
pages = "51033--51051",
ISBN = "979-8-89176-462-0",
abstract = "As large language models are deployed at scale, the ability to selectively remove knowledge has become critical for privacy, safety, and legal compliance. Existing post-training unlearning methods suffer from a persistent tension: unbounded divergence signals over-unlearn, degrading utility and producing incoherent outputs, while fixed sample-agnostic targets and reference-relative penalties under-unlearn, leaving traces recoverable under paraphrase or membership-inference attacks. We attribute both failures to a common structural cause: the forgetting objective is decoupled from any sample- or parameter-level notion of what retention requires. We present FLOUR (Forgetting via Learned On-manifold Unlearning Representations), a latent unlearning framework in which a lightweight encoder produces a per-sample suppression direction. This direction is anchored to the null region of the reference model{'}s activations, keeping it on-manifold and disentangled from retention-critical directions, which enables forgetting while preserving coherent generation and enhancing privacy. Results on the Open-Unlearning benchmark suite show that FLOUR most closely matches the retrain oracle across the unlearning trilemma of forgetting, retention, and privacy, substantially outperforming prior baselines. Our code is publicly available at https://github.com/ judydnguyen/flour-unlearning."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nguyen-etal-2026-flour">
<titleInfo>
<title>FLOUR: Mixing the Right Ingredients for LLM Unlearning in Latent Space</title>
</titleInfo>
<name type="personal">
<namePart type="given">Thuy</namePart>
<namePart type="given">Dung</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Giang</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Taylor</namePart>
<namePart type="given">T</namePart>
<namePart type="family">Johnson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kevin</namePart>
<namePart type="family">Leach</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>As large language models are deployed at scale, the ability to selectively remove knowledge has become critical for privacy, safety, and legal compliance. Existing post-training unlearning methods suffer from a persistent tension: unbounded divergence signals over-unlearn, degrading utility and producing incoherent outputs, while fixed sample-agnostic targets and reference-relative penalties under-unlearn, leaving traces recoverable under paraphrase or membership-inference attacks. We attribute both failures to a common structural cause: the forgetting objective is decoupled from any sample- or parameter-level notion of what retention requires. We present FLOUR (Forgetting via Learned On-manifold Unlearning Representations), a latent unlearning framework in which a lightweight encoder produces a per-sample suppression direction. This direction is anchored to the null region of the reference model’s activations, keeping it on-manifold and disentangled from retention-critical directions, which enables forgetting while preserving coherent generation and enhancing privacy. Results on the Open-Unlearning benchmark suite show that FLOUR most closely matches the retrain oracle across the unlearning trilemma of forgetting, retention, and privacy, substantially outperforming prior baselines. Our code is publicly available at https://github.com/ judydnguyen/flour-unlearning.</abstract>
<identifier type="citekey">nguyen-etal-2026-flour</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2376</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2376/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>51033</start>
<end>51051</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T FLOUR: Mixing the Right Ingredients for LLM Unlearning in Latent Space
%A Nguyen, Thuy Dung
%A Nguyen, Giang
%A Johnson, Taylor T.
%A Leach, Kevin
%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 nguyen-etal-2026-flour
%X As large language models are deployed at scale, the ability to selectively remove knowledge has become critical for privacy, safety, and legal compliance. Existing post-training unlearning methods suffer from a persistent tension: unbounded divergence signals over-unlearn, degrading utility and producing incoherent outputs, while fixed sample-agnostic targets and reference-relative penalties under-unlearn, leaving traces recoverable under paraphrase or membership-inference attacks. We attribute both failures to a common structural cause: the forgetting objective is decoupled from any sample- or parameter-level notion of what retention requires. We present FLOUR (Forgetting via Learned On-manifold Unlearning Representations), a latent unlearning framework in which a lightweight encoder produces a per-sample suppression direction. This direction is anchored to the null region of the reference model’s activations, keeping it on-manifold and disentangled from retention-critical directions, which enables forgetting while preserving coherent generation and enhancing privacy. Results on the Open-Unlearning benchmark suite show that FLOUR most closely matches the retrain oracle across the unlearning trilemma of forgetting, retention, and privacy, substantially outperforming prior baselines. Our code is publicly available at https://github.com/ judydnguyen/flour-unlearning.
%R 10.18653/2026.findings-emnlp.2376
%U https://aclanthology.org/2026.findings-emnlp.2376/
%U https://doi.org/10.18653/2026.findings-emnlp.2376
%P 51033-51051
Markdown (Informal)
[FLOUR: Mixing the Right Ingredients for LLM Unlearning in Latent Space](https://aclanthology.org/2026.findings-emnlp.2376/) (Nguyen et al., Findings 2026)
ACL