@inproceedings{zhao-etal-2026-p2p,
title = "{P}2{P}: A Poison-to-Poison Remedy for Reliable Backdoor Defense in {LLM}s",
author = "Zhao, Shuai and
Wu, Xinyi and
Zhao, Shiqian and
Wu, Xiaobao and
Guo, Zhongliang and
Jia, Yanhao and
Luu, Anh Tuan",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.600/",
pages = "12345--12360",
ISBN = "979-8-89176-395-1",
abstract = "Defending Large Language Models (LLMs) against backdoor attacks has long been trapped in a ``cat-and-mouse'' dilemma, where defenders passively react to ever-shifting attack strategies. To break this cycle, we posit that proactive immunization is inherently superior to reactive sanitization. In this study, we propose Poison-to-Poison (P2P), a general and effective defense algorithm that introduces a paradigm shift. Instead of waiting to detect malicious samples, P2P strategically implants benign triggers to reshape the model{'}s decision boundary, redirecting latent feature activation from malicious trajectories to a safe, controllable output space. This enforces the model to associate trigger-induced representations with safe outputs, thereby overriding the effects of original malicious triggers. Thanks to this robust and generalizable trigger-based fine-tuning, P2P is effective across task settings and attack types. Theoretically and empirically, we show that P2P can neutralize malicious backdoors while preserving task performance. We conduct extensive experiments on classification, mathematical reasoning, and summary generation tasks, involving multiple state-of-the-art LLMs. The results demonstrate that our P2P algorithm significantly reduces the attack success rate compared with baseline models. We hope that P2P can serve as a practical guideline for defending against backdoor attacks in the Model as a Service (MaaS) scenario, where benign prompts are embedded within the system to regulate model behavior."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhao-etal-2026-p2p">
<titleInfo>
<title>P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shuai</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinyi</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shiqian</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaobao</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhongliang</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanhao</namePart>
<namePart type="family">Jia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anh</namePart>
<namePart type="given">Tuan</namePart>
<namePart type="family">Luu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Defending Large Language Models (LLMs) against backdoor attacks has long been trapped in a “cat-and-mouse” dilemma, where defenders passively react to ever-shifting attack strategies. To break this cycle, we posit that proactive immunization is inherently superior to reactive sanitization. In this study, we propose Poison-to-Poison (P2P), a general and effective defense algorithm that introduces a paradigm shift. Instead of waiting to detect malicious samples, P2P strategically implants benign triggers to reshape the model’s decision boundary, redirecting latent feature activation from malicious trajectories to a safe, controllable output space. This enforces the model to associate trigger-induced representations with safe outputs, thereby overriding the effects of original malicious triggers. Thanks to this robust and generalizable trigger-based fine-tuning, P2P is effective across task settings and attack types. Theoretically and empirically, we show that P2P can neutralize malicious backdoors while preserving task performance. We conduct extensive experiments on classification, mathematical reasoning, and summary generation tasks, involving multiple state-of-the-art LLMs. The results demonstrate that our P2P algorithm significantly reduces the attack success rate compared with baseline models. We hope that P2P can serve as a practical guideline for defending against backdoor attacks in the Model as a Service (MaaS) scenario, where benign prompts are embedded within the system to regulate model behavior.</abstract>
<identifier type="citekey">zhao-etal-2026-p2p</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.600/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>12345</start>
<end>12360</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs
%A Zhao, Shuai
%A Wu, Xinyi
%A Zhao, Shiqian
%A Wu, Xiaobao
%A Guo, Zhongliang
%A Jia, Yanhao
%A Luu, Anh Tuan
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F zhao-etal-2026-p2p
%X Defending Large Language Models (LLMs) against backdoor attacks has long been trapped in a “cat-and-mouse” dilemma, where defenders passively react to ever-shifting attack strategies. To break this cycle, we posit that proactive immunization is inherently superior to reactive sanitization. In this study, we propose Poison-to-Poison (P2P), a general and effective defense algorithm that introduces a paradigm shift. Instead of waiting to detect malicious samples, P2P strategically implants benign triggers to reshape the model’s decision boundary, redirecting latent feature activation from malicious trajectories to a safe, controllable output space. This enforces the model to associate trigger-induced representations with safe outputs, thereby overriding the effects of original malicious triggers. Thanks to this robust and generalizable trigger-based fine-tuning, P2P is effective across task settings and attack types. Theoretically and empirically, we show that P2P can neutralize malicious backdoors while preserving task performance. We conduct extensive experiments on classification, mathematical reasoning, and summary generation tasks, involving multiple state-of-the-art LLMs. The results demonstrate that our P2P algorithm significantly reduces the attack success rate compared with baseline models. We hope that P2P can serve as a practical guideline for defending against backdoor attacks in the Model as a Service (MaaS) scenario, where benign prompts are embedded within the system to regulate model behavior.
%U https://aclanthology.org/2026.findings-acl.600/
%P 12345-12360
Markdown (Informal)
[P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs](https://aclanthology.org/2026.findings-acl.600/) (Zhao et al., Findings 2026)
ACL
- Shuai Zhao, Xinyi Wu, Shiqian Zhao, Xiaobao Wu, Zhongliang Guo, Yanhao Jia, and Anh Tuan Luu. 2026. P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs. In Findings of the Association for Computational Linguistics: ACL 2026, pages 12345–12360, San Diego, California, United States. Association for Computational Linguistics.