@inproceedings{kowalczyk-etal-2026-data,
title = "Data-Driven Red-Teaming: Reusing Adversarial Attacks via Metadata-Aware Retrieval",
author = "Kowalczyk, Mateusz and
Sendera, Marcin and
Cygert, Sebastian",
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.1622/",
doi = "10.18653/2026.findings-emnlp.1622",
pages = "34510--34528",
ISBN = "979-8-89176-462-0",
abstract = "As large language models (LLMs) are increasingly deployed in real-world systems, red-teaming has become a critical tool for identifying safety vulnerabilities. However, existing approaches remain largely stateless and inefficient, repeatedly rediscovering similar failure modes for each new model. In this work, we propose a data-driven paradigm for red-teaming that enables the accumulation and reuse of adversarial knowledge across models. We introduce a framework for generating and collecting adversarial attacks enriched with structured metadata, including attack strategies, risk categories, and diagnostic feedback signals. Leveraging this resource, we demonstrate that metadata-aware retrieval significantly improves the efficiency of discovering successful attacks on unseen models, reducing reliance on purely generative search. To support further research, we release a large-scale dataset comprising {\ensuremath{>}}188k successful attacks across 8 models, together with rich metadata annotations."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kowalczyk-etal-2026-data">
<titleInfo>
<title>Data-Driven Red-Teaming: Reusing Adversarial Attacks via Metadata-Aware Retrieval</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mateusz</namePart>
<namePart type="family">Kowalczyk</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marcin</namePart>
<namePart type="family">Sendera</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Cygert</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 (LLMs) are increasingly deployed in real-world systems, red-teaming has become a critical tool for identifying safety vulnerabilities. However, existing approaches remain largely stateless and inefficient, repeatedly rediscovering similar failure modes for each new model. In this work, we propose a data-driven paradigm for red-teaming that enables the accumulation and reuse of adversarial knowledge across models. We introduce a framework for generating and collecting adversarial attacks enriched with structured metadata, including attack strategies, risk categories, and diagnostic feedback signals. Leveraging this resource, we demonstrate that metadata-aware retrieval significantly improves the efficiency of discovering successful attacks on unseen models, reducing reliance on purely generative search. To support further research, we release a large-scale dataset comprising \ensuremath>188k successful attacks across 8 models, together with rich metadata annotations.</abstract>
<identifier type="citekey">kowalczyk-etal-2026-data</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.1622</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1622/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>34510</start>
<end>34528</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Data-Driven Red-Teaming: Reusing Adversarial Attacks via Metadata-Aware Retrieval
%A Kowalczyk, Mateusz
%A Sendera, Marcin
%A Cygert, Sebastian
%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 kowalczyk-etal-2026-data
%X As large language models (LLMs) are increasingly deployed in real-world systems, red-teaming has become a critical tool for identifying safety vulnerabilities. However, existing approaches remain largely stateless and inefficient, repeatedly rediscovering similar failure modes for each new model. In this work, we propose a data-driven paradigm for red-teaming that enables the accumulation and reuse of adversarial knowledge across models. We introduce a framework for generating and collecting adversarial attacks enriched with structured metadata, including attack strategies, risk categories, and diagnostic feedback signals. Leveraging this resource, we demonstrate that metadata-aware retrieval significantly improves the efficiency of discovering successful attacks on unseen models, reducing reliance on purely generative search. To support further research, we release a large-scale dataset comprising \ensuremath>188k successful attacks across 8 models, together with rich metadata annotations.
%R 10.18653/2026.findings-emnlp.1622
%U https://aclanthology.org/2026.findings-emnlp.1622/
%U https://doi.org/10.18653/2026.findings-emnlp.1622
%P 34510-34528
Markdown (Informal)
[Data-Driven Red-Teaming: Reusing Adversarial Attacks via Metadata-Aware Retrieval](https://aclanthology.org/2026.findings-emnlp.1622/) (Kowalczyk et al., Findings 2026)
ACL