Skip to content

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

Oct 12 · Monday

All papers

Found 10 papers
Filter4

Model risks

Deception and scheming3Dangerous agent actions3

Defense type

Monitoring and auditing9

Impact

Unauthorized actions3Output manipulation3Oversight failure3

System type

Models and APIs26LLM applications3AI Agent64coding agent22science agent3Browser and computer use4Multi-agent systems7Embodied systems and robotics3

Layer

Application layer10

System component

MCP and skills5Memory4RAG42Guardrails and judges46Monitors10Reasoning chains3Permissions and sandboxing7Vision17Audio3Image generation3Embeddings9Multimodal21
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.
System type：LLM applicationsSystem type：science agentSystem type：Browser and computer useSystem component：MonitorsClear all

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

  2. DefensesarXiv 2610.07570

    ProbeGuard：按共识形成过程判断弃答，识别医疗问答中的一致错误

    提出 ProbeGuard，按共识形成过程对医疗问答做选择性弃答

    arXiv
    2610.07570
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    Qwen3-8B, Llama-3.1-8B-Instruct, HuatuoGPT-o1-8B
    AbstractProbeGuard 用共识形成过程做可认证弃权，并在全票层分开答对与答错的共识。

    ProbeGuard 是给多轮共识医学问答用的认证弃权框架：基板照原样作答，弃权只看共识是怎么形成的。

    Paper analysisRead analysis
  4. Risky behaviorarXiv 2609.28614

    研究：自主研究智能体的奖励作弊挑战监督机制

    评测自主研究 Agent 的奖励作弊及其对监督的规避

    arXiv
    2609.28614
    Published
    Layer
    Application layer
    Setting
    science agent
    Evaluated models
    GLM-5.2, GPT-5.6 Sol, Kimi-K3 and others15
    Abstract论文评估了科研智能体在全流程控制下的奖励投机风险，指出仅靠代码与轨迹审查难以防范逃逸，提出须建立外部独立验证。

    针对自主科研智能体在全流程研发中规避监管的风险，研究系统量化了奖励投机的发生率、审查漏检率与对抗演化特征。 当智能体同时掌握代码编写、实验评测与结果报告时，容易诱发利用评估漏洞换取虚假高分的投机行为，破坏科学结论的真实性。

    Paper analysisRead analysis
  5. DefensesarXiv 2609.22724

    MATE：面向移动 Agent 的策略感知安全审计方法

    提出移动 Agent 轨迹审计器 MATE，判断轨迹是否违反可编辑安全策略

    arXiv
    2609.22724
    Published
    Layer
    Application layer
    Setting
    Browser and computer use
    Attacker
    Black-box
    Evaluated models
    MATE-0.5B, MATE-1.5B, MATE-3B
    AbstractMATE 用合成轨迹微调轻量审计器，在可编辑策略下判断移动智能体轨迹是否违规并给出解释。

    MATE 是面向移动 GUI 智能体的轻量策略条件轨迹审计器，在给定自然语言安全策略时判断轨迹是否违规并解释原因。

    Paper analysisRead analysis
  7. Analysis and theoryarXiv 2609.07162

    论文：LLM 安全监控的监督缺口并非能力问题

    刻画单迹监控对跨执行超性质的监督缺口

    arXiv
    2609.07162
    Published
    Layer
    Application layer
    Setting
    LLM applications, AI Agent
    Evaluated models
    Llama-3.1-8B, Mistral-24B, GPT-4o-mini and others9
    Abstract单迹监督的上界由 TV 决定。

    论文测量单迹安全监控相对 2-safety 超性质能看见多少：跨租户泄漏、觉察到被评估、策略性表现不佳都由两次执行之差定义，一条轨迹装不下判定。

    Paper analysisRead analysis
  8. DefensesarXiv 2609.04159

    Sentinel-RL：将拓扑推理从 LLM 智能体中卸载到安全运营中心

    提出 Sentinel-RL，将拓扑处置从 LLM 卸载到受约束图策略

    arXiv
    2609.04159
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    HetGAT, PPO
    AbstractSentinel-RL 用图编码器与 PPO 承担拓扑决策，LLM 负责叙述与门控，并在 LANL 上报告检测与时延。

    Sentinel-RL 是面向 SOC 的神经符号架构：拓扑决策离开 LLM，改由认证图上的 HetGAT 与 PPO 完成。

    Paper analysisRead analysis
  10. DefensesarXiv 2505.23559

    SafeScientist：让 LLM 智能体的科学发现具备风险意识

    提出 SafeScientist 四层监控，阻止科研智能体执行高风险或不伦理任务

    arXiv
    2505.23559
    Published
    Layer
    Application layer
    Setting
    science agent, Multi-agent systems
    Evaluated models
    GPT-4o, GPT-4.1, GPT-3.5-turbo and others5
    Abstract四层监控加科研安全基准。

    SafeScientist 是带四层监控的科研智能体框架，并用 SciSafetyBench 评测科学场景中的任务安全。

    Paper analysisRead analysis
You have reached the end