Zhicheng Fang, Han Sun, Jiachen Lei, Xinyi Fang, Zongmin Zhang, Chenxu Fu, Sizhe Fang, Wei Xu · 2026 · arXiv
DOI: 10.48550/arXiv.2610.25897 · open-access PDF
A medida que los ataques de jailbreak contra los grandes modelos de lenguaje (LLM) se vuelven cada vez más diversos, el red-teaming ha llegado a depender de grandes bibliotecas de promesas de jailbreak y estrategias de ataque para evaluar la robustez de los modelos de víctimas y sus configuraciones de defensa. Sin embargo, la eficacia del ataque es altamente condicional: un ataque que tenga éxito bajo una configuración de defensa de víctimas puede fallar bajo otra, mientras que la evaluación exhaustiva de la biblioteca de ataque completa para cada configuración incurre en costes de token y monetarios sustanciales.
Machine translation (Meta M2M100); the English original follows.
As jailbreak attacks against large language models (LLMs) become increasingly diverse, red-teaming has come to rely on large libraries of jailbreak prompts and attack strategies to evaluate the robustness of victim models and their defense settings. However, attack effectiveness is highly condition-dependent: an attack that succeeds under one victim-defense setting may fail under another, while exhaustively evaluating the full attack library for each setting incurs substantial token and monetary costs. This motivates studying how jailbreak attacks should be prioritized under a limited attempt
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.25897, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange