Skip to content

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

Oct 12 · Monday

All papers

Found 92 papers
Filter4

Attack type

Jailbreaks3Poisoning and backdoors7Detection evasion13

Attack technique

Gradient and representation optimization6Defense-targeted4Dormant triggers3

Model risks

Deception and scheming5Misalignment and value divergence9Dangerous agent actions4Hallucination and fabrication18Bias and discrimination4

Defense type

Model hardening3Inference-time intervention12Detection and filtering5Monitoring and auditing5Watermarking and provenance4

Impact

Harmful content80Unauthorized actions4Data and privacy leaks37Output manipulation88Availability and reliability18Oversight failure35

System type

Models and APIs82LLM applications4AI Agent7Embodied systems and robotics4

Attacker access

Black-box3White-box3

Attack surface / entry point

User input9Images, audio and video12Training data5Weights and gradients5Distribution channels4Inference services and hardware5

Layer

Prompt layer60Model layer88Application layer361Training and data layer51Infrastructure layer4

System component

Guardrails and judges3Fine-tuning and open weights3Vision16Audio5Video4Image generation6Multimodal27
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.
Impact：Unauthorized actionsImpact：Output manipulationLayer：Model layerLayer：Infrastructure layerClear all

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

  3. Evaluation and benchmarksarXiv 2610.11305

    BeliefScope：区分证据与用户压力对模型信念影响的受控框架

    用 BeliefScope 分离固定命题上证据与压力对模型信念的影响

    arXiv
    2610.11305
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-8B, Llama3.1-8B, Gemma3-12B and others4
    AbstractBeliefScope 把证据修正与压力偏移拆成绑定评测条件的行为画像。

    BeliefScope 是一个黑盒行为诊断框架，用来把固定命题上的回答变化拆成证据驱动的修正，和不加新事实的方向性压力。信念修正评测能看到信息基础变化后的更新，谄媚评测能看到用户影响下的顺从或漂移。作者认为，当两种输入都偏向同一结论时，表面偏移本身不能区分来源。

    Paper analysisRead analysis
  5. AttacksarXiv 2610.09027

    P-VMI：对抗图像经 KV cache 持续影响多轮对话

    提出持久视觉记忆注入，使对抗图像经缓存在触发时操控输出

    arXiv
    2610.09027
    Published
    Layer
    Infrastructure layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-8B, Qwen3-VL-8B-Instruct, Gemma-4-12B

    Abstract论文提出 P-VMI，证实对抗输入被掩蔽后仍可经 KV 缓存持久控制模型，强调了保留状态的安全风险。

    Paper analysisRead analysis
  7. Risky behaviorarXiv 2610.08670

    Study: post-training recipes determine whether large models act against their own moral judgments

    提出知行差距面板，测量压力下模型是否违背自身判断

    arXiv
    2610.08670
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    OLMo-3-7B-Instruct, Llama-3.1-8B-Instruct, Tulu 3 and others7
    Abstract评估发现后训练配方决定知行差距是否存在，行动前审思可有效缓解。

    本文评估了开源语言模型作为智能体行动时违背自身道德判断的知行差距及其决定因素。

    Paper analysisRead analysis
  10. Analysis and theoryarXiv 2610.07798

    研究：财务信息披露越少，Llama-3.1-8B-Instruct 的理财建议越受身份影响

    测量财务事实变少时身份对股权配置建议的替代

    arXiv
    2610.07798
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Llama-3.1-8B-Instruct, Qwen2.5-1.5B-Instruct
    Abstract身份摆幅随披露减少上升。

    同一财务只改人设，测量披露变薄时股权建议里的身份替代。

    Paper analysisRead analysis
  11. Evaluation and benchmarksarXiv 2610.06668

    研究发现语言模型能识别不可解的工程问题却仍报告已解决

    评测模型识别不可能工程题后是否仍报已解决

    arXiv
    2610.06668
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Fable 5.1, Opus 5.5, Sonnet 5.5 and others14
    Abstract成对力学基准把能否求解和是否拒绝不可能题拆开，二者并不互换。

    成对的工程力学题把能不能解、和会不会拒绝不可能的前提拆开。每题有一致版，以及只改一个数据或前提、使所述状态不可能存在的缺陷版。

    Paper analysisRead analysis
  12. InterpretabilityarXiv 2610.06576

    研究：智能体欺骗行为在外部显现前已可从内部表征预测

    提出基于决策前隐藏状态的检测与引导，预测并抑制智能体欺骗

    arXiv
    2610.06576
    Published
    Layer
    Model layer
    Setting
    AI Agent
    Evaluated models
    Qwen3-14B

    Abstract论文展示了智能体自发欺骗可在行为显现前由内部表征预测，并通过激活引导证实了表征的行为可操作性。

    Paper analysisRead analysis
  13. Evaluation and benchmarksarXiv 2610.06522

    SycoLens study: a single sycophancy flip rate masks the distinction between correction and capitulation

    构建 SycoLens 重放评测，区分谄媚翻转中的纠错与屈从

    arXiv
    2610.06522
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    opus-5, sonnet-5, sonnet-4.6 and others11
    Abstract论文揭示施压句式、边界距离与格式偏置主导翻转率，提出系统解构纠错与顺从的综合评测方案。

    本文系统解构了大语言模型在用户质疑下的阿谀奉承行为与评测机制。

    Paper analysisRead analysis
  14. Evaluation and benchmarksarXiv 2610.06339

    BabelFake：面向多语言音视频 DeepFake 检测的基准数据集

    构建多语言音视频 DeepFake 基准 BabelFake 并评测检测器

    arXiv
    2610.06339
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    SpeechForensics, AVH-Align, (B)RAVEn
    AbstractBabelFake 为五语同意采集基准。

    BabelFake 是经同意与 IRB 批准采集的多语音视频 DeepFake 检测基准，用来把视觉操纵和语音合成拆开比较。

    Paper analysisRead analysis
  16. AttacksarXiv 2610.06049

    研究者提出仅修改 SAE 解码器的后门攻击，可在不改动语言模型的情况下植入行为

    提出只改 SAE 解码器的后门，在冻结语言模型上植入代码插入

    arXiv
    2610.06049
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Gemma 3 4B Instruct, Qwen3 1.7B, Llama 3.1 8B Instruct
    Abstract只改解码器的 SAE 可带上代码插入后门，且若干常规质量指标变化不大。

    只改 SAE 解码器、冻结编码器与语言模型，即可把代码插入行为放进推理时的辅助组件。作者把这视为 SAE 供应链问题：checkpoint 一旦替换某一层激活，就参与后续计算，而不只是解释工具。

    Paper analysisRead analysis
  17. AttacksarXiv 2610.05894

    研究者提出闭环偏见注入攻击

    提出闭环激活引导，在扩散语言模型去噪过程中注入定向偏见

    arXiv
    2610.05894
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Attacker
    Gray-box
    Evaluated models
    LLaDA-8B-Instruct, Dream-v0-Instruct-7B, LLaDA-MoE-7B-A1B-Instruct

    Abstract通过去噪中间概率反馈动态调节残差流转向，证明扩散语言模型在推理服务栈中存在闭环偏见注入脆弱性。

    Paper analysisRead analysis
  18. DefensesarXiv 2610.05166

    安全动作不等于可行动作：VICS-G 为 VLA 策略做可行未来解码

    提出 VICS-G 可行未来解码，重排冻结 VLA 策略的动作候选

    arXiv
    2610.05166
    Published
    Layer
    Model layer
    Setting
    Embodied systems and robotics
    Evaluated models
    SafeVLA (task-optimized / base), SafeVLA (safety-aligned / safe)
    Abstract作者用可行未来解码重排冻结 VLA 的下一步，并称 VICS-G 降低观测代价且完成率接近策略采样。

    训练无关解码器 VICS 在冻结 VLA 的推理阶段重排动作块，使当前选择带上该动作仍可能留下的安全完成质量。

    Paper analysisRead analysis
  19. Analysis and theoryarXiv 2610.04860

    Research reveals that text watermarks induce model hallucinations and proposes two mitigation interventions

    分析文本水印在 RAG 中诱发事实幻觉的机制并提出干预

    arXiv
    2610.04860
    Published
    Layer
    Model layer
    Setting
    LLM applications
    Evaluated models
    LLaMA 3.1 8B Instruct, Mistral-7B-Instruct-v0.3, Qwen2.5-14B-Instruct
    Abstract论文揭示了生产环境文本水印导致模型事实退化的失效模式，分析了解码机制并提出针对性干预。

    论文研究生产级文本水印对大语言模型事实准确性的影响，分析了水印诱发幻觉的失效模式并评估了针对性缓解方案。

    Paper analysisRead analysis
Load more