Skip to content

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

Oct 12 · Monday

All papers

Found 3 papers
Filter5

Attack type

Detection evasion3

Attack technique

Gradient and representation optimization24Automated search3Encoding and obfuscation3Role-play and persuasion4Conversation structure manipulation0Cross-modal payloads14Defense-targeted7Dormant triggers3

Impact

Harmful content48Output manipulation3

Attack surface / entry point

User input3

Layer

Prompt layer3Model layer0Application layer9
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.
Attack technique：Encoding and obfuscationAttack technique：Conversation structure manipulationImpact：Output manipulationLayer：Prompt layerLayer：Model layerClear all

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

  1. Evaluation and benchmarksarXiv 2609.05843

    SinoGlyphBench：中文字形混淆下语言模型审核的诊断基准

    提出 SinoGlyphBench，用配对字形混淆诊断中文审核稳定性

    arXiv
    2609.05843
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    GPT-5.4 mini, GPT-5.5, DeepSeek V4 Flash and others12
    AbstractSinoGlyphBench 显示中文字形混淆会同时增加有害漏报和误报，全文扰动最重。

    SinoGlyphBench是一个配对诊断基准，用来看内容审核在可还原的中文字形混淆下是否还保持原决定。

    Paper analysisRead analysis
  2. AttacksarXiv 2606.09700

    The researchers propose the HPAA attack: bypassing 13 content moderation systems with typographic visual cues

    提出 HPAA 攻击，用排版视觉线索绕过内容审核系统

    arXiv
    2606.09700
    Published
    Layer
    Prompt layer
    Setting
    LLM applications
    Attacker
    Black-box
    Evaluated models
    Llama-Guard-3-8B, Perspective API, ShieldGemma-2B and others13
    Abstract揭示了排版视觉线索与分词差异带来的审核盲区，HPAA 以极低查询实现近乎完全规避。

    这篇论文研究了人类视觉感知与大模型分词处理机制之间的差异所导致的内容审核漏洞。

    Paper analysisRead analysis
You have reached the end