Skip to content

Translation in progress Some entries still include text in the original language.

Oct 12 · Monday

All papers

Found 21 papers
Filter4
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 233papers have no tags yet and are not included in filtered results.

Another 233papers do not have category tags yet and are not included in filtered results.

  1. AttacksarXiv 2609.03247

    研究揭示大语言模型的自发身份认证失败:五款模型的分阶段开发者身份实验

    提出 MIPC 与 CFA,观察模型自拟测验后接受开发者身份

    Published
    Evaluated models
    ChatGPT, Claude, Qwen3.7-Plus and others5
    Abstract自拟测验可造成对话层虚假认证,但所测授权边界并未因此改变。

    这项工作把“模型自己出题、自己批改、再宣布身份已核验”定义成对话层认证失败,并与后端提权分开。

    Paper analysisRead analysis
  2. AttacksarXiv 2609.02414

    Blueprint 框架用世界观模拟与 18 个影响因子提升多轮越狱成功率

    提出多轮越狱框架 BLUEPRINT,用世界观模拟与心理因素诱导有害输出

    Published
    Attacker
    Black-box
    Evaluated models
    Qwen3-Next-80B, DeepSeek-V3.2, GLM-4.7 and others6
    AbstractBLUEPRINT 解耦多轮越狱策略与情境模拟,揭示任务具象化是摆脱拒绝的核心机制,为多轮安全防御提供新视角。

    BLUEPRINT 是一个将因子化社会影响力策略空间与跨轮世界观模拟解耦的多轮大语言模型安全评测框架。

    Paper analysisRead analysis
  3. AttacksarXiv 2607.23496

    Concept2Scenario:用 SAE 概念归因发现可迁移的越狱场景

    提出 Concept2Scenario,借助内部概念归因发现可迁移的越狱场景

    Published
    Evaluated models
    Qwen3.5-9B, Llama-3.1-8B, Ministral-3-8B and others6

    Abstract该研究从表征空间因果归因揭示场景削弱拒绝的机制,发现可复用脆弱场景先验,提升多模型黑盒越狱效能与探索速度。

    Paper analysisRead analysis
  4. MultiTurnPSB

    构建 MultiTurnPSB,评测四轮医疗越狱与分类器输入防御

    Published
    Evaluated models
    GPT-4.1-mini, Claude Sonnet 4.5
    Abstract多轮攻击加剧医疗模型风险,分类器干预有效但受高误报率制约。

    MultiTurnPSB 针对患者端医疗大模型在持续交互中的安全风险,构建了包含三种攻击模态的四轮对抗评测基准。

    Paper analysisRead analysis
  5. AttacksarXiv 2604.04842

    First, do no harm: revealing hidden vulnerabilities in LLMs for psychological counseling via persona-based client simulation attacks

    Propose PCSA, which uses persona-based multi-turn simulation of counseling clients to induce harmful empathetic responses.

    Published
    Evaluated models
    Llama-3.1-8B, Llama-3.1-70B, GPT-3.5-Turbo and others8
    Paper analysisRead analysis
  6. Cooking Risks: Benchmarking and Mitigating Food Safety Risks in Large Language Models

    Propose FoodGuardBench and the guardrail FoodGuard-4B to evaluate and reduce food hazard jailbreaks

    Published
    Attacker
    Black-box
    Evaluated models
    Claude-3.7-Sonnet, GPT-4o, GPT-4.1 and others10
    Paper analysisRead analysis
  7. AttacksarXiv 2602.00388

    Safer by Diffusion, Broken by Context: The Safety Dividend and Failure Modes of Diffusion LLMs

    提出黑盒 context nesting,用结构化补全任务嵌套有害请求越狱扩散 LLM

    Published
    Attacker
    Black-box
    Evaluated models
    LLaDA-8B-Instruct, LLaDA-1.5, Dream-v0-Instruct-7B and others5
    Abstract去噪抑制多种越狱。

    作者把 D-LLM 相对 AR-LLM 越狱的稳健性称为 safety blessing,并把 context nesting 当作它的 black-box 失效方式。

    Paper analysisRead analysis
  8. AttacksarXiv 2601.22169

    In Vino Veritas and Vulnerabilities:用醉酒语言诱导检验 LLM 安全

    用醉酒语言诱导改编对话模型并评测越狱与隐私泄露

    Published
    Evaluated models
    LLaMA2-7B, LLaMA3-8B, Mistral-7B and others5
    Abstract醉酒语言诱导在五个 LLM 上抬高越狱与部分隐私错误,作者将其视为攻击向量。

    作者把醉酒语言当作攻击向量,检验把 LLM 适配成醉酒发短信的模式后,越狱和情境隐私是否变差。动机是人类醉酒时的失范与泄密,以及 LLM 模拟人设的能力。

    Paper analysisRead analysis