@inproceedings{fu-etal-2026-self,
title = "Self-Sanitize: Cognitive-Inspired In-Place Mitigation of {LLM} Harmful Responses",
author = "Fu, Wenjie and
Wang, Huandong and
Gao, Chen and
Wan, Guoan and
Gao, Junyao and
Li, Yong and
Jiang, Tao",
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.1390/",
pages = "31732--31757",
ISBN = "979-8-89176-459-0",
abstract = "Despite significant advances in aligning Large Language Models (LLMs) with safety and ethical standards, adversarial prompts can still be crafted to elicit undesirable responses. Existing mitigation strategies predominantly rely on post-hoc filtering, which introduces substantial latency or computational overhead, and is incompatible with token-level streaming generation. To address this, we introduce Self-Sanitize, a novel LLM-driven mitigation framework inspired by cognitive psychology, which emulates human self-monitor and self-repair behaviors. The framework utilizes a lightweight Self-Monitor module to inspect high-level intentions via representation engineering and a Self-Repair module for in-place correction without separate review dialogues. This design allows for real-time streaming monitoring and seamless repair, with negligible impact on latency and resource utilization. Extensive experiments across three privacy-leakage scenarios and broader safety-sensitive domains (e.g., crime and bias) demonstrate that Self-Sanitize achieves superior mitigation performance with minimal overhead and preserved model utility. Our code is available at the following link: https://anonymous.4open.science/r/LLM{\_}Self{\_}Sanitize-5E8B"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="fu-etal-2026-self">
<titleInfo>
<title>Self-Sanitize: Cognitive-Inspired In-Place Mitigation of LLM Harmful Responses</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wenjie</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Huandong</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chen</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guoan</namePart>
<namePart type="family">Wan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junyao</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yong</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tao</namePart>
<namePart type="family">Jiang</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>Despite significant advances in aligning Large Language Models (LLMs) with safety and ethical standards, adversarial prompts can still be crafted to elicit undesirable responses. Existing mitigation strategies predominantly rely on post-hoc filtering, which introduces substantial latency or computational overhead, and is incompatible with token-level streaming generation. To address this, we introduce Self-Sanitize, a novel LLM-driven mitigation framework inspired by cognitive psychology, which emulates human self-monitor and self-repair behaviors. The framework utilizes a lightweight Self-Monitor module to inspect high-level intentions via representation engineering and a Self-Repair module for in-place correction without separate review dialogues. This design allows for real-time streaming monitoring and seamless repair, with negligible impact on latency and resource utilization. Extensive experiments across three privacy-leakage scenarios and broader safety-sensitive domains (e.g., crime and bias) demonstrate that Self-Sanitize achieves superior mitigation performance with minimal overhead and preserved model utility. Our code is available at the following link: https://anonymous.4open.science/r/LLM_Self_Sanitize-5E8B</abstract>
<identifier type="citekey">fu-etal-2026-self</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1390/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>31732</start>
<end>31757</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Self-Sanitize: Cognitive-Inspired In-Place Mitigation of LLM Harmful Responses
%A Fu, Wenjie
%A Wang, Huandong
%A Gao, Chen
%A Wan, Guoan
%A Gao, Junyao
%A Li, Yong
%A Jiang, Tao
%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 fu-etal-2026-self
%X Despite significant advances in aligning Large Language Models (LLMs) with safety and ethical standards, adversarial prompts can still be crafted to elicit undesirable responses. Existing mitigation strategies predominantly rely on post-hoc filtering, which introduces substantial latency or computational overhead, and is incompatible with token-level streaming generation. To address this, we introduce Self-Sanitize, a novel LLM-driven mitigation framework inspired by cognitive psychology, which emulates human self-monitor and self-repair behaviors. The framework utilizes a lightweight Self-Monitor module to inspect high-level intentions via representation engineering and a Self-Repair module for in-place correction without separate review dialogues. This design allows for real-time streaming monitoring and seamless repair, with negligible impact on latency and resource utilization. Extensive experiments across three privacy-leakage scenarios and broader safety-sensitive domains (e.g., crime and bias) demonstrate that Self-Sanitize achieves superior mitigation performance with minimal overhead and preserved model utility. Our code is available at the following link: https://anonymous.4open.science/r/LLM_Self_Sanitize-5E8B
%U https://aclanthology.org/2026.emnlp-main.1390/
%P 31732-31757
Markdown (Informal)
[Self-Sanitize: Cognitive-Inspired In-Place Mitigation of LLM Harmful Responses](https://aclanthology.org/2026.emnlp-main.1390/) (Fu et al., EMNLP 2026)
ACL