多篇研究论文探讨了增强大型语言模型（LLM）安全性、使其免受越狱攻击的先进技术。这些研究引入了新的框架和方法，用于评估和防御旨在诱导有害输出的对抗性提示。研究重点在于开发更全面的评估指标、自适应攻击生成策略以及能够识别模型行为中细微模式的鲁棒检测机制。 AI
影响 LLM安全性和越狱检测方面的进展对于在敏感应用中负责任地部署AI至关重要。
排序理由 该集群包含多篇在arXiv上发表的学术论文，详细介绍了关于LLM安全性和越狱攻击的新研究。
AI 生成摘要 · Google Gemini · 来自 8 个来源。 我们如何撰写摘要 →
arXiv:2605.06605v1 Announce Type: new Abstract: Evaluating and predicting the performance of large language models (LLMs) in multi-turn conversational settings is critical yet computationally expensive; key events -- e.g., jailbreaks or successful task completion by an agent -- o…
Evaluating and predicting the performance of large language models (LLMs) in multi-turn conversational settings is critical yet computationally expensive; key events -- e.g., jailbreaks or successful task completion by an agent -- often emerge only after repeated interactions. Th…
Large Language Models (LLMs) have achieved remarkable success but remain highly susceptible to jailbreak attacks, in which adversarial prompts coerce models into generating harmful, unethical, or policy-violating outputs. Such attacks pose real-world risks, eroding safety, trust,…
arXiv:2605.02958v1 Announce Type: cross Abstract: Representation Engineering typically relies on static refusal vectors derived from terminal representations. We move beyond this paradigm, demonstrating that refusal is a dynamic and sparse process rather than a localized outcome.…
arXiv:2605.02921v1 Announce Type: cross Abstract: As LLMs continue to shape real-world applications, automated jailbreak generation becomes essential to reveal safety weaknesses and guide model improvement. Existing automatic jailbreak generation methods have not yet fully consid…
arXiv:2605.01687v1 Announce Type: new Abstract: We present MultiBreak, a scalable and diverse multi-turn jailbreak benchmark to evaluate large language model (LLM) safety. Multi-turn jailbreaks mimic natural conversational settings, making them easier to bypass safety-aligned LLM…
arXiv:2601.19487v2 Announce Type: replace Abstract: Safety-aligned LLMs suffer from two failure modes: jailbreak (answering harmful inputs) and over-refusal (declining benign queries). Existing vector steering methods adjust the magnitude of answer vectors, but this creates a fun…
arXiv:2605.00974v1 Announce Type: cross Abstract: LLMs are increasingly equipped with safety alignment mechanisms, yet recent studies demonstrate that they remain vulnerable to jailbreaking attacks that elicit harmful behaviors without explicit policy violations. While a growing …