@inproceedings{han-etal-2026-leveraging,
title = "Leveraging Teacher-Forcing Token-Level Anomaly Signals for Heterogeneous Backdoor Filtering in {LLM}s",
author = "Han, Tianxu and
Wang, Jiangming and
Zhou, Chunyi and
Zhao, Binbin and
Du, Tianyu",
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.585/",
pages = "12351--12370",
ISBN = "979-8-89176-462-0",
abstract = "Generative backdoor attacks pose a growing threat to the instruction tuning of large language models (LLMs), where attackers poison a few training samples to induce malicious payloads on triggered inputs while preserving normal behavior on clean ones. However, existing backdoor-poisoned sample filtering methods typically rely on global sample representations and clustering-based separation, and such global signals can become fragile when heterogeneous target-side malicious payloads scatter poisoned samples or low poisoning rates prevent poisoned samples from forming stable clusters. This motivates us to move beyond cross-sample separability and instead exploit localized token-level target-side anomalies induced by malicious payload tokens under teacher forcing. Building on this intuition, we propose TTAF, a Teacher-forcing Token-level Anomaly Filtering framework which introduces a token-level anomaly score under teacher forcing and aggregates localized evidence to filter suspicious poisoned samples before fine-tuning. Experiments across multiple datasets, representative backdoor attacks, and different LLMs show that TTAF consistently achieves strong filtering performance within this payload-injection setting, with efficient runtime and no external dependencies. Code is available at https://github.com/Sam-Han2020/TTAF."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="han-etal-2026-leveraging">
<titleInfo>
<title>Leveraging Teacher-Forcing Token-Level Anomaly Signals for Heterogeneous Backdoor Filtering in LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tianxu</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiangming</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chunyi</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Binbin</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tianyu</namePart>
<namePart type="family">Du</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>Generative backdoor attacks pose a growing threat to the instruction tuning of large language models (LLMs), where attackers poison a few training samples to induce malicious payloads on triggered inputs while preserving normal behavior on clean ones. However, existing backdoor-poisoned sample filtering methods typically rely on global sample representations and clustering-based separation, and such global signals can become fragile when heterogeneous target-side malicious payloads scatter poisoned samples or low poisoning rates prevent poisoned samples from forming stable clusters. This motivates us to move beyond cross-sample separability and instead exploit localized token-level target-side anomalies induced by malicious payload tokens under teacher forcing. Building on this intuition, we propose TTAF, a Teacher-forcing Token-level Anomaly Filtering framework which introduces a token-level anomaly score under teacher forcing and aggregates localized evidence to filter suspicious poisoned samples before fine-tuning. Experiments across multiple datasets, representative backdoor attacks, and different LLMs show that TTAF consistently achieves strong filtering performance within this payload-injection setting, with efficient runtime and no external dependencies. Code is available at https://github.com/Sam-Han2020/TTAF.</abstract>
<identifier type="citekey">han-etal-2026-leveraging</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.585/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>12351</start>
<end>12370</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Leveraging Teacher-Forcing Token-Level Anomaly Signals for Heterogeneous Backdoor Filtering in LLMs
%A Han, Tianxu
%A Wang, Jiangming
%A Zhou, Chunyi
%A Zhao, Binbin
%A Du, Tianyu
%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 han-etal-2026-leveraging
%X Generative backdoor attacks pose a growing threat to the instruction tuning of large language models (LLMs), where attackers poison a few training samples to induce malicious payloads on triggered inputs while preserving normal behavior on clean ones. However, existing backdoor-poisoned sample filtering methods typically rely on global sample representations and clustering-based separation, and such global signals can become fragile when heterogeneous target-side malicious payloads scatter poisoned samples or low poisoning rates prevent poisoned samples from forming stable clusters. This motivates us to move beyond cross-sample separability and instead exploit localized token-level target-side anomalies induced by malicious payload tokens under teacher forcing. Building on this intuition, we propose TTAF, a Teacher-forcing Token-level Anomaly Filtering framework which introduces a token-level anomaly score under teacher forcing and aggregates localized evidence to filter suspicious poisoned samples before fine-tuning. Experiments across multiple datasets, representative backdoor attacks, and different LLMs show that TTAF consistently achieves strong filtering performance within this payload-injection setting, with efficient runtime and no external dependencies. Code is available at https://github.com/Sam-Han2020/TTAF.
%U https://aclanthology.org/2026.findings-emnlp.585/
%P 12351-12370
Markdown (Informal)
[Leveraging Teacher-Forcing Token-Level Anomaly Signals for Heterogeneous Backdoor Filtering in LLMs](https://aclanthology.org/2026.findings-emnlp.585/) (Han et al., Findings 2026)
ACL