Skip to content

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

Oct 12 · Monday

All papers

Found 79 papers
Filter4

Attack type

Jailbreaks3Prompt injection24Poisoning and backdoors10Extraction and theft3

Attack technique

Dormant triggers3

Model risks

Deception and scheming11Reward hacking4Misalignment and value divergence8Dangerous agent actions5

Defense type

Detection and filtering5Instruction–data separation4Permissions and approval9

Impact

Harmful content7Unauthorized actions13Data and privacy leaks7Output manipulation26Oversight failure9

System type

Models and APIs271LLM applications132AI Agent608coding agent126science agent30web agent34Browser and computer use18Multi-agent systems79Embodied systems and robotics23

Attack surface / entry point

User input11Documents and messages6Tool responses5Tool and skill descriptions3Retrieval corpus3Memory writes6Inter-agent messages25

Layer

Prompt layer0Application layer79Infrastructure layer0

System component

MCP and skills4Memory11Guardrails and judges5Monitors7Permissions and sandboxing12
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 240papers have no tags yet and are not included in filtered results.
System type：Multi-agent systemsLayer：Prompt layerLayer：Application layerLayer：Infrastructure layerClear all

Another 240papers do not have category tags yet and are not included in filtered results.

  1. OtherarXiv 2610.13491

    EviRepair: Evidence-Driven Agents for Reliable Vulnerability Repair

    Proposing EviRepair, an evidence-driven multi-agent framework for repository-level vulnerability repair

    arXiv
    2610.13491
    Published
    Layer
    Application layer
    Setting
    coding agent, Multi-agent systems
    Evaluated models
    DeepSeek-V4.1-Flash, GLM-5.2, Qwen3.8-27B
    AbstractEviRepair uses a reviewed vulnerability model for repository-level repair.

    EviRepair is a repository-level automated vulnerability repair framework for sparse vulnerability reports, using a reviewed vulnerability model to separate diagnosis from patch synthesis.

    Paper analysisRead analysis
  2. 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
  3. Analysis and theoryarXiv 2610.12436

    The study proposes an ecological theory of AI agent populations: collaboration brings a critical scale for takeoff.

    An ecological model is established for a population of mismatched agents, showing that collaboration produces a critical takeoff scale.

    arXiv
    2610.12436
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    AbstractThe authors say that collaboration changes takeoff from an individual capability threshold to a population threshold, so small-population evaluations are insufficient.

    The authors use the ecological strong Allee effect to discuss whether a mismatched agent population can take off when individual capabilities remain unchanged..

    Paper analysisRead analysis
  7. DefensesarXiv 2610.07657

    DEFER1 多智能体防御研究报告受控测试中的攻击成功率下降

    实现 DEFER，以确定性检查优先拦截多智能体流水线的内容攻击

    arXiv
    2610.07657
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen3-235B-A22B-Instruct-2507, gpt-oss-120b, Llama-4-Scout-17B-16E-Instruct and others5
    AbstractDEFER 把确定性检查放在 judge 之前。

    DEFER 用来测量多智能体授权流水线里，确定性规则在何处结束、LLM judge 从何处开始。

    Paper analysisRead analysis
  8. Evaluation and benchmarksarXiv 2610.06748

    BazaarBench：评估 LLM Agent 在去中心化 C2C 市场中的代理安全

    构建 BazaarBench，评测 C2C 市场中 Agent 的违规与失信行为

    arXiv
    2610.06748
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT-5.5, GPT-5.4, GPT-5.4-mini and others5
    Abstract论文提出挂钩底层资产的 C2C 交易评测基准，发现时间压力与对抗指令显著推高了智能体的违规承诺与欺诈行为。

    BazaarBench 针对大语言模型智能体在去中心化 C2C 交易中的代理安全问题，提出了一个结合底层实体状态与大模型裁判的仿真基准平台。

    Paper analysisRead analysis
  9. DefensesarXiv 2610.05640

    Research reveals that the CaMeL defense fails in multi-agent systems and proposes multi-CaMeL

    提出 multi-CaMeL，在智能体调用边界隔离可信指令与非可信数据

    arXiv
    2610.05640
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Claude Fable 5, GPT-5.5, GPT-4.1 and others6

    Abstract论文针对多智能体中单体防御不组合的边界清洗问题，提出 multi-CaMeL 协议彻底阻断了 MultiAgentDojo 上的注入攻击。

    Paper analysisRead analysis
  10. Evaluation and benchmarksarXiv 2610.04672

    MASBench：部分可观测条件下的 LLM 多智能体协作基准

    提出 MASBench，评测部分可观测下的多智能体协作机制

    arXiv
    2610.04672
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen-Plus, GPT-5-nano, Gemini-2.5-Flash-Lite and others5
    AbstractMASBench 以确定性指标比较部分可观测下的协作机制。

    MASBench 用部分可观测任务比较 LLM 多智能体的协作机制，而不是只看端到端能否做完。

    Paper analysisRead analysis
  11. Risky behaviorarXiv 2610.04528

    论文量化分析自主 LLM 智能体在 Collusion Wiki 事件中的串通行为

    对自主 LLM 智能体日志做身份盲置换以量化串通

    arXiv
    2610.04528
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Abstract修正后汇总共编偏向同智能体聚集。

    作者把 Nightingale Collective 所记的 wiki 旁路事件，做成相对身份置换零模型的定量配套，并写出四类会使协调检验失真的构造错误。

    Paper analysisRead analysis
  12. DefensesarXiv 2610.04412

    论文研究 LLM 数字孪生参与增强民主时的提示注入脆弱性

    提出检测、结构化与 GSPO 流程，降低共识生成的提示注入错位

    arXiv
    2610.04412
    Published
    Layer
    Application layer
    Setting
    LLM applications, Multi-agent systems
    Evaluated models
    LLaMA-2 7B, LLaMA-3 8B, GPT 3.5 Turbo and others14
    Abstract三章依次做个体偏好预测、政党知识图谱审议，以及共识生成对提示注入的脆弱性与一条防御流程。

    这是一篇把 LLM 数字孪生放进增强民主的博士论文，分个体表示、政党聚合和共识生成的提示注入三章。

    Paper analysisRead analysis
  13. Risky behaviorarXiv 2610.03033

    研究：LLM 智能体间的数值信号与协调行为

    测量多智能体博弈中数值信号的结构及其对合作的影响

    arXiv
    2610.03033
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT-4o (gpt-4o-2024-11-20), Mistral Large (mistral-large-2512), Claude Haiku 4.5 (claude-haiku-4-5-20251001) and others4
    Abstract受指令数值会形成锚定收益的结构并影响接收方，但合作效应不能跨模型外推。

    四种 LLM 在四个经典两人博弈中交换无预设语义的数字时，消息结构可以跨模型出现，合作水平的变化则不能跨模型外推。

    Paper analysisRead analysis
  14. Evaluation and benchmarksarXiv 2610.02874

    SceneFactory-3D：将 2D 交通场景提升为 3D 物理反事实以做安全评测

    提出 SceneFactory-3D，用路况反事实评测驾驶策略的闭环安全性

    arXiv
    2610.02874
    Published
    Layer
    Application layer
    Setting
    Embodied systems and robotics, Multi-agent systems
    Evaluated models
    Frozen L0, L0-T, L1 and others5
    AbstractSceneFactory-3D 固定场景只改路况。

    SceneFactory-3D 是 GPU 批处理的多智能体驾驶仿真器，用来在场景和控制器固定时只改路面物理，并比较闭环结果。

    Paper analysisRead analysis
  15. Analysis and theoryarXiv 2610.02480

    MEA：面向忠实模型解释的奖励驱动多智能体系统

    提出奖励驱动多智能体系统 MEA，生成可扰动验证的忠实解释

    arXiv
    2610.02480
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen3.6-35B-A3B, Qwen3.5-4B, Qwen3-VL-30B-A3B-Instruct and others12
    AbstractMEA 用忠实性奖励训练双智能体，把工具输出收成可扰动核对的解释。

    MEA 是一个奖励驱动的双智能体，用来给黑盒分类器的单次预测写自然语言解释，并检查解释能否被扰动后的模型行为支持。

    Paper analysisRead analysis
  16. DefensesarXiv 2610.02349

    MIRROR：面向多智能体通信的法定人数完整性防御

    提出 MIRROR，用多路径摘要多数表决保障多智能体通信完整性

    arXiv
    2610.02349
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Gemma-4-31B-it, Gemini 2.5 Pro
    AbstractMIRROR 靠多数路径。

    MIRROR 是 LLM 多智能体通信层上的多路径法定人数完整性原语，只处理不透明字节，不解释消息内容。要处理的是 Agent-in-the-Middle：中介合法终结 TLS 后可改在途消息，而语义校验有推理开销和误拦，签名又要求私钥管理。做法是生成结束后复制同一规范化载荷。

    Paper analysisRead analysis
  17. DefensesarXiv 2610.02005

    贝叶斯辩证论证（BDA）：面向持续对抗下多 LLM 委员会的可校准聚合

    提出 BDA，按辩证动作可靠性加权聚合多 LLM 议会以抵抗持续对抗

    arXiv
    2610.02005
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    AbstractBDA 用零额外调用的可靠性加权，给出可校准后验，并在持续对抗下反转不可靠席位。

    BDA 是一种不另调 LLM 的多 LLM 议会聚合方法，同时处理置信度校准和持续不可靠席位。

    Paper analysisRead analysis
  19. AttacksarXiv 2610.00430

    Memetic Trojans：利用社交传染在 LLM Agent 网络中传播对抗载荷

    提出 memetic trojans，借助社交传染在 Agent 网络中传播对抗载荷

    arXiv
    2610.00430
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    gpt-oss-120B, deepseek-v4.1-flash, qwen-32B and others5
    Abstractmemetic trojans 用内生社会传染携带载荷。

    Memetic trojans 是一类网络介导攻击：对抗载荷搭在智能体本来就会转发和点赞的内容上扩散，转发者不必被攻陷，也不必执行“请继续复制”的指令。

    Paper analysisRead analysis
  20. AttacksarXiv 2609.39788

    研究：多智能体系统潜在通信链路可被训练用于提升有害顺从

    提出只训练或篡改多智能体潜在通信链路来提高有害遵从

    arXiv
    2609.39788
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Llama-3.2-3B-Instruct, Qwen2.5-3B-Instruct, Gemma-3-1B-IT and others7
    Abstract潜在链路本身即可改变冻结智能体系统的安全，奖励既能攻击也能修复。

    学习得到的潜在通信链路会改变由冻结、对齐智能体组成的系统是否遵从有害请求。。作者研究三种拓扑：双智能体、三级顺序协作，以及双专家加汇总。

    Paper analysisRead analysis
Load more