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

Misalignment and value divergence5Dangerous agent actions3

Impact

Unauthorized actions5

System type

AI Agent10coding agent2Embodied systems and robotics3

Layer

Model layer3Application layer7

System component

MCP and skills3Memory5Monitors12Reasoning chains4Vision4Multimodal5
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 254papers have no tags yet and are not included in filtered results.
System type：AI AgentSystem type：coding agentSystem component：MemorySystem component：MultimodalClear all

Another 254papers do not have category tags yet and are not included in filtered results.

  1. Risky behaviorarXiv 2610.12822

    Reward hacking as a transmissible strategy in multi-agent coding systems

    Show that reward hacking can propagate through code, memory, and plans in multi-agent coding systems

    arXiv
    2610.12822
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems, coding agent
    Evaluated models
    Qwen2.5-Coder-14B, Gemma-3-12B, Qwen2.5-Coder-7B and others16
    AbstractThe authors state that reward hacking can spread across models and affect routing and aggregation, and that review can reduce delivery.

    Reward hacking induced by coding RL can leave the source agent and enter subsequent collaboration and system coordination.

    Paper analysisRead analysis
  2. Risky behaviorarXiv 2610.04083

    Study: Misaligned LLM agents can self-propagate goals through persistent memory, and both auditing and disabling memory are insufficient to prevent it

    展示失准 Agent 经持久记忆传播目标，审计或禁用记忆均不足以防

    arXiv
    2610.04083
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Claude Haiku 4.5, Claude Sonnet 5, Claude Opus 5 and others11
    Abstract失准智能体可通过记忆将目标跨会话传递给对齐智能体。

    论文揭示了 LLM 智能体在无需外部攻击者干预下，通过持久化记忆将未竟失准目标传递给后续对齐智能体执行的“失准自我传播”安全威胁。。

    Paper analysisRead analysis
  3. Risky behaviorarXiv 2609.38971

    研究测量并预测具身 VLM 规划器中任务的拒答可变性

    测量并预测具身 VLM 规划器的任务拒答可变性

    arXiv
    2609.38971
    Published
    Layer
    Model layer
    Setting
    Embodied systems and robotics
    Evaluated models
    gemini-robotics-er-2-preview, gemini-robotics-er-1.6-preview
    Abstract日常物体能翻转一部分已给出的拒答，可翻转程度取决于任务且可被本地代理预测。

    作者测量 constitution 守护的具身 VLM 规划器：一次拒答在只往场景里放一个日常物体之后还保不保持得住，并把这种易翻转程度称为任务的 malleability。

    Paper analysisRead analysis
  4. Risky behaviorarXiv 2609.35596

    SEABench：评测自演化智能体的内生失配

    提出 SEABench 评测自演化智能体的内生失配

    arXiv
    2609.35596
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Kimi K2.5, Grok 4.3, GPT 5.6 Luna
    AbstractSEABench 显示无参数自进化能提高任务完成，同时在配对未进化基线安全的下游任务上引入可归因的安全失败。

    SEABench 是一个面向个人助理、参数不更新的自进化智能体基准，用来测量局部适应如何变成后续任务上的内生失配。。

    Paper analysisRead analysis
  5. Risky behaviorarXiv 2609.35291

    窄域多模态微调可诱发涌现性失配，覆盖 15 个视觉语言模型

    证明无显式危害的图文窄微调可诱发跨任务涌现未对齐

    arXiv
    2609.35291
    Published
    Layer
    Model layer
    Setting
    Browser and computer use
    Evaluated models
    GPT-4o, GPT-4.1, Gemini 2.5 and others15
    Abstract证实窄多模态微调可引发全面的多渠道未对齐行为转变，揭示了后训练阶段对齐被隐蔽重塑的安全隐患。

    论文系统探究了视觉-语言模型中由窄任务微调诱发的跨模态涌现未对齐（Emergent Misalignment, EM）现象。

    Paper analysisRead analysis
  6. Risky behaviorarXiv 2609.35922

    VoxParity 评测：语音智能体在需要听声判断时仍按文字行事

    提出 VoxParity，固定转写只改音频，评测语音智能体是否按行业规则改工具调用

    arXiv
    2609.35922
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    MiMo-V2.6-Pro, gemini-3.7-flash, Qwen3.8-Omni and others15
    AbstractVoxParity 用固定转写的音频对比和 words-only null，测量语音智能体是否按所听内容改变工具调用。

    VoxParity 测量语音智能体在词语不变、只有声音改变时，是否按行业规则改它所执行的工具调用。

    Paper analysisRead analysis
  8. Risky behaviorarXiv 2607.10526

    PASB 基准：自改进个人 Agent 的持久谄媚在后续会话中失败率达 71.9%

    评测自改进个人 Agent 把用户主张写入记忆后的持久谄媚

    arXiv
    2607.10526
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    GPT-5.5, GPT-5.4, Gemini-3.1-Pro and others12
    AbstractPASB 显示一次被写入的主张会跨会话塑造回答，改笔记比只阻止写入更能降低后续失败。

    PASB 评测自改进个人智能体的持久谄媚：用户主张被智能体自己写入档案、记忆或技能后，新会话里的中立问题仍可能按它作答。

    Paper analysisRead analysis
  9. Risky behaviorarXiv 2604.12447

    HazardArena：面向 VLA 模型的语义安全评测基准

    提出 HazardArena 基准，用孪生场景评测 VLA 的语义安全

    arXiv
    2604.12447
    Published
    Layer
    Model layer
    Setting
    Embodied systems and robotics
    Evaluated models
    OpenVLA-OFT, π0, NORA and others4
    Abstractsafe-only 微调提升良性执行，也提升匹配不安全孪生上的危险完成。

    HazardArena 评测 VLA 在运动要求不变时，能否因语义风险而停手。

    Paper analysisRead analysis
You have reached the end