AgenticRed: Evolving Agentic Systems for Red-Teaming

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yuan, Jiayi, Nöther, Jonathan, Jaques, Natasha, Radanović, Goran
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914441477488640
author Yuan, Jiayi
Nöther, Jonathan
Jaques, Natasha
Radanović, Goran
author_facet Yuan, Jiayi
Nöther, Jonathan
Jaques, Natasha
Radanović, Goran
contents While recent automated red-teaming methods show promise for systematically exposing model vulnerabilities, most existing approaches rely on human-specified workflows. This dependence on manually designed workflows suffers from human biases and makes exploring the broader design space expensive. We introduce AgenticRed, an automated pipeline that leverages LLMs' in-context learning to iteratively design and refine red-teaming systems without human intervention. Rather than optimizing attacker policies within predefined structures, AgenticRed treats red-teaming as a system design problem, and it autonomously evolves automated red-teaming systems using evolutionary selection and generational knowledge. Red-teaming systems designed by AgenticRed consistently outperform state-of-the-art approaches, achieving 96% attack success rate (ASR) on Llama-2-7B, 98% on Llama-3-8B and 100% on Qwen3-8B on HarmBench. Our approach generates robust, query-agnostic red-teaming systems that transfer strongly to the latest proprietary models, achieving an impressive 100% ASR on GPT-5.1, DeepSeek-R1 and DeepSeek V3.2. This work highlights evolutionary algorithms as a powerful approach to AI safety that can keep pace with rapidly evolving models.
format Preprint
id arxiv_https___arxiv_org_abs_2601_13518
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle AgenticRed: Evolving Agentic Systems for Red-Teaming
Yuan, Jiayi
Nöther, Jonathan
Jaques, Natasha
Radanović, Goran
Artificial Intelligence
Neural and Evolutionary Computing
While recent automated red-teaming methods show promise for systematically exposing model vulnerabilities, most existing approaches rely on human-specified workflows. This dependence on manually designed workflows suffers from human biases and makes exploring the broader design space expensive. We introduce AgenticRed, an automated pipeline that leverages LLMs' in-context learning to iteratively design and refine red-teaming systems without human intervention. Rather than optimizing attacker policies within predefined structures, AgenticRed treats red-teaming as a system design problem, and it autonomously evolves automated red-teaming systems using evolutionary selection and generational knowledge. Red-teaming systems designed by AgenticRed consistently outperform state-of-the-art approaches, achieving 96% attack success rate (ASR) on Llama-2-7B, 98% on Llama-3-8B and 100% on Qwen3-8B on HarmBench. Our approach generates robust, query-agnostic red-teaming systems that transfer strongly to the latest proprietary models, achieving an impressive 100% ASR on GPT-5.1, DeepSeek-R1 and DeepSeek V3.2. This work highlights evolutionary algorithms as a powerful approach to AI safety that can keep pace with rapidly evolving models.
title AgenticRed: Evolving Agentic Systems for Red-Teaming
topic Artificial Intelligence
Neural and Evolutionary Computing
url https://arxiv.org/abs/2601.13518