Skip to content

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

Oct 12 · Monday

All papers

Found 230 papers
Filter3

Attack type

Jailbreaks178Model tampering14Detection evasion4Malicious use5

Attack technique

Gradient and representation optimization20Automated search40Multi-turn escalation20Encoding and obfuscation35Role-play and persuasion16Conversation structure manipulation9Reasoning-chain manipulation6Cross-modal payloads30Defense-targeted5

Model risks

Misalignment and value divergence4Inappropriate refusal35Bias and discrimination4

Defense type

Model hardening8Inference-time intervention19Detection and filtering18Instruction–data separation7Monitoring and auditing8

Impact

Harmful content230Unauthorized actions6Data and privacy leaks39Output manipulation142Availability and reliability25Oversight failure40

System type

Models and APIs222AI Agent7coding agent3

Attacker access

Black-box87Gray-box4White-box27

Attack surface / entry point

User input172Images, audio and video44Training data3Weights and gradients16

Layer

Prompt layer152Model layer78Application layer30Training and data layer49

System component

Guardrails and judges45Monitors9Reasoning chains12Fine-tuning and open weights7Vision34Audio10Image generation21Multimodal62
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 265papers have no tags yet and are not included in filtered results.
Impact：Harmful contentLayer：Prompt layerLayer：Model layerClear all

Another 265papers do not have category tags yet and are not included in filtered results.

  1. DefensesarXiv 2610.13733

    KVeil: First-Place Solution for the ECCV 2026 LifeGenIP Unlearnable Video Challenge

    Propose KVeil, which adds perturbations to face videos to prevent unauthorized customization from reproducing identity

    arXiv
    2610.13733
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    White-box, Black-box
    Evaluated models
    Wan2.2-TI2V-5B, HunyuanVideo-1.5, Kling 3.0 and others4

    AbstractKVeil adapts LETHE's reference-reading disruption to the challenge's two stages, ranking first on the leaderboard in both Final columns.

    Paper analysisRead analysis
  2. DefensesarXiv 2610.13333

    LETHE: Defending Against Unauthorized Image-to-Video Generation via Value Diversity Collapse

    Proposes LETHE, which perturbs reference images to block unauthorized image-to-video

    arXiv
    2610.13333
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Wan2.1-I2V-14B, CogVideoX-5B-I2V, LightX2V and others6
    AbstractLETHE suppresses the diversity of reference values and attracts attention, making protected images difficult to animate faithfully.

    LETHE is a protective perturbation added to personal images before publication, used to counter unauthorized image-to-video generation.

    Paper analysisRead analysis
  3. InterpretabilityarXiv 2610.13078

    Attention sees, MLP refuses: tracing the safety pathway from attention to MLP in LLMs

    Trace attention to identify harmfulness to the MLP refusal write pathway and propose Sentinel

    arXiv
    2610.13078
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, Qwen3-8B-Instruct and others18
    AbstractAttention updates are more readable for harmfulness, MLP updates write refusal, and Sentinel routes alerts using two-path signals.

    The authors reframe jailbreaking and excessive refusal of aligned models as a pathway problem on the residual stream from recognition to writing, and accordingly provide conditional reminders before decoding.

    Paper analysisRead analysis
  4. Evaluation and benchmarksarXiv 2610.11766

    研究提出意图恢复率作为 LLM 安全评测的补充指标

    提出 IRIS，用意图恢复率区分遮蔽越狱下无害输出的原因

    arXiv
    2610.11766
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    GPT-4o, Claude Sonnet 4.5, Claude Opus 4.5 and others6
    Abstract无害输出的证据权重取决于任务是否被恢复。

    IRIS 把意图恢复当作中间信号，用来解释意图遮蔽提示下为何没有出现有害协助。

    Paper analysisRead analysis
  5. DefensesarXiv 2610.09703

    研究揭示 Token-Pruning 对 VLM 安全性的影响并提出 SAP 缓解机制

    提出 Safety-Aware Pruning，缓解视觉 token 剪枝放大的多模态越狱

    arXiv
    2610.09703
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    LLaVA-1.5-7B
    Abstract剪枝去掉背景 token 后注意力塌向恶意前景。

    这篇工作评估视觉 token 剪枝如何改变 VLM 在多模态越狱下的行为，并给出推理时防御 SAP。

    Paper analysisRead analysis
  6. DefensesarXiv 2610.09167

    Persist-3 用顺序检验识别未见越狱来源，提升 MTK 与 Prompt Guard 2 防御效果

    提出 Persist-3 序贯检验，从隐藏表示识别未见越狱源

    arXiv
    2610.09167
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Llama2-7B-chat, Qwen2.5-7B-Instruct, Gemma-2-9B-it and others4
    AbstractPersist-3 用有限窗口检验攻击流是否超出已知越狱源，检出率低于全历史 SR-DKW，但决策只依赖最近观测。

    Persist-3 是一个有限记忆的序贯检测器，用来判断攻击表示流是否已经不能由任何一个已知越狱源解释。。

    Paper analysisRead analysis
  7. Evaluation and benchmarksarXiv 2610.09033

    研究发布 ASRD 数据集，评测五款开源模型在非规范输入下的安全表现

    arXiv
    2610.09033
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Gemma 2 9B, Gemma 3 4B, Llama 3.1 8B and others6
    Abstract五模型、七族表面形式的四态评测里，低有害遵从不等于读懂后的安全处理。

    ASRD 用四态标签检查五个开源模型如何处理非规范表面上的有害请求。

    Paper analysisRead analysis
  9. DefensesarXiv 2610.08678

    研究揭示投机解码的安全代价并提出 SecureSD 加固方法

    测量有损投机解码的安全代价，并提出 SecureSD 收紧早期验证

    arXiv
    2610.08678
    Published
    Layer
    Model layer
    Setting
    AI Agent
    Attacker
    Black-box
    Evaluated models
    Qwen3-8B, Qwen3-32B, Llama-3.1-8B-Instruct and others6
    Abstract有损推测解码的早期偏离更快抬高越狱与注入风险。

    作者研究有损推测解码在换取速度时，是否会比效用更快地损害越狱与提示注入防护，并给出按位置校正的 SecureSD。

    Paper analysisRead analysis
  10. DefensesarXiv 2610.07774

    研究者提出读取路由 logits 的 MoE 视觉语言模型多模态安全检测器

    AI summary研究者提出一种轻量级路由 logits 安全检测器，在提示词预填充阶段读取 MoE 视觉语言模型的专家路由信号，在生成前判断多模态请求是否安全，且不修改模型参数或专家路由。

    arXiv
    2610.07774
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-VL-30B-A3B-Instruct, Kimi-VL-A3B-Instruct
    Abstract读出而非操纵 router logits，可在不改模型的情况下降低两模型的组合式安全错误。

    作者在两个指令微调 MoE VLM 上比较改模型与读路由。组合式风险按 HoliSafe 分成五类：图文各自可安全，合在一起才不安全，或不安全内容只在图像里。

    Paper analysisRead analysis
  11. DefensesarXiv 2610.07532

    LADE：用暗知识中的模型无关潜在安全信号防御 LLM 越狱

    提出 LADE，用首 token 暗知识在生成前过滤越狱查询

    arXiv
    2610.07532
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    White-box
    Evaluated models
    Llama-2-7B-Chat, Llama-3-8B-Instruct, Qwen2-7B-Instruct and others13
    Abstract作者用 LADE 在生成前过滤有害查询，并称八个模型中六个的 Held-out 平均最好。

    LADE 是一种只读首 token 输出概率的防御，用来在生成前挡下可能引出有害回复的查询。

    Paper analysisRead analysis
  12. AttacksarXiv 2610.07125

    PEV 攻击利用嵌入向量加噪越狱六款开源权重模型

    提出 PEV 攻击，在输入嵌入上加高斯噪声越狱开源模型

    arXiv
    2610.07125
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Llama-3.1-8B-Instruct, Qwen2.5-1.5B-Instruct, Phi-4-mini-3.8B and others6
    AbstractPEV 用 embedding 高斯噪声，在六个模型上越狱 JailbreakBench 全部提示。

    PEV 是对 open-weight LLM 的 embedding 噪声越狱。

    Paper analysisRead analysis
  13. AttacksarXiv 2610.06670

    研究者提出 Reward Stealing Attack，从对齐模型行为反推安全奖励实施攻击

    提出 ReSA，从对齐模型行为恢复安全奖励并在解码时反转

    arXiv
    2610.06670
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Attacker
    White-box, Gray-box
    Evaluated models
    Llama3.1-8B-Instruct, Gemma-7B-Instruct, Qwen2.5-7B-Instruct and others9
    AbstractReSA 用最大熵 IRL 从对齐模型行为恢复代理安全奖励，并在解码时反转以诱导有害输出。

    ReSA 是一种对抗攻击：从对齐模型的可观察行为恢复代理安全奖励，再在解码时把该奖励反过来用。。

    Paper analysisRead analysis
  15. DefensesarXiv 2610.06383

    基于影响力的曲率感知数据加权激活引导方法

    提出曲率感知数据加权的激活转向方法，用于毒性抑制与概念控制

    arXiv
    2610.06383
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Gemma-2-2B, Llama-3.2-3B
    Abstract影响加权传输压低毒性率，闭式 IF-Act 常换来更高困惑度与更低 MMLU。

    推理时的影响加权激活传输，用曲率修正的样本权重替换传输转向里的均匀类均值。要做的是概念控制，而不是再训练。作者认为类均值被无关概念、噪声和少数 token 主导，编码的是词元级而非主题。

    Paper analysisRead analysis
  16. InterpretabilityarXiv 2610.06064

    TrustMI：通过模型激活因果控制助手对用户的信任

    提出 TrustMI，用激活转向因果调控助手对用户主张的信任

    arXiv
    2610.06064
    Published
    Layer
    Model layer
    Setting
    AI Agent
    Evaluated models
    Qwen3.5-9B, Qwen3.5-27B, Llama-3.1-8B-Instruct and others6
    Abstract作者用 BiPO 转向调控助手对用户的依赖，并称该效应迁移到智能体安全任务。

    TrustMI 研究助手对用户的信任能否沿激活中的线性方向被因果调控，以及这种调控会不会改变安全相关的智能体行为。

    Paper analysisRead analysis
  17. InterpretabilityarXiv 2610.05541

    研究者提出 CAP 指标与 CSFD 算法，发现越狱通过压制安全特征绕过拒答

    提出 CAP 指标与 CSFD 算法，发现被抑制的安全特征

    arXiv
    2610.05541
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Gemma-2-2B, Qwen3-8B, Gemma-3-4B-IT and others5

    Abstract论文指出越狱通过抑制安全特征运作，提出 CAP 与 CSFD 补充了针对激活特征的解释性方法。

    Paper analysisRead analysis
  18. DefensesarXiv 2610.05346

    研究者提出序列马赛克攻击并给出防御理论

    形式化 watchman 防御以阻止序贯马赛克攻击

    arXiv
    2610.05346
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3.6-35B-A3B, Gemma-4-26B-A4B-it, DeepSeek-V2-Lite (16B-A2.4B)
    Abstractwatchman 理论给出零失败且有用的条件。

    这篇工作同时给出序贯马赛克防御的博弈理论，以及冻结 LLM 上的分角色自博弈训练。

    Paper analysisRead analysis
Load more