Skip to content

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

Oct 13 · Tuesday

All papers

Found 7 papers
Filter6

Attack type

Jailbreaks6Prompt injection9Extraction and theft1Detection evasion6

Attack technique

Gradient and representation optimization5Automated search8Multi-turn escalation5Encoding and obfuscation14Conversation structure manipulation2Reasoning-chain manipulation8Cross-modal payloads28

Impact

Harmful content6

System type

Models and APIs7

Attacker access

Black-box6White-box3

Attack surface / entry point

User input5Images, audio and video4

Layer

Prompt layer7

System component

Guardrails and judges3Reasoning chains3Vision4Multimodal6
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 213papers have no tags yet and are not included in filtered results.
Attack type：JailbreaksAttack type：Extraction and theftAttack technique：Gradient and representation optimizationAttack technique：Conversation structure manipulationSystem component：Reasoning chainsSystem component：VisionClear all

Another 213papers do not have category tags yet and are not included in filtered results.

  3. AttacksarXiv 2608.09867

    Researchers steal proprietary LLM inference chains via cross-model compatibility of encrypted inference blocks

    利用加密思维块的跨模型重放，诱导弱模型逐字转写专有推理链

    arXiv
    2608.09867
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5 and others13

    Abstract论文揭示了客户端加密思维块跨模型互换漏洞，提出了低成本提取方法并给出防御方案。

    Paper analysisRead analysis
  4. AttacksarXiv 2605.07399

    研究者提出 GPO-V，用全局概率优化越狱扩散视觉语言模型

    提出 GPO-V，用全局概率优化的视觉扰动越狱扩散视觉语言模型

    arXiv
    2605.07399
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    White-box, Black-box
    Evaluated models
    LLaDA-V, LaViDA
    Abstract作者用视觉扰动改 dVLM 的全局生成概率，以绕过两类拒绝。

    GPO-V 把越狱从“开头必须顺从”改成“整段去噪概率朝肯定方向走”。。

    Paper analysisRead analysis
  6. AttacksarXiv 2511.16110

    MFA 框架在 17 个视觉语言模型上取得 58.5% 攻击成功率

    提出 MFA 框架，协同突破视觉语言模型的对齐、提示与审核层

    arXiv
    2511.16110
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    MiniGPT-4, LLaMA-4-Scout-Inst, LLaMA-3.2-11B-Vision-Instruct and others17

    Abstract作者称多层防御的 VLM 仍存在跨模型漏洞，MFA 整合三项技术实现了端到端越狱。

    Paper analysisRead analysis
You have reached the end