Skip to content

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

Oct 13 · Tuesday

All papers

Found 21 papers
Filter5

Model risks

Deception and scheming12Reward hacking4Misalignment and value divergence9Dangerous agent actions5

Impact

Output manipulation8Oversight failure6

System type

Models and APIs39AI Agent56coding agent12science agent3Multi-agent systems21

Layer

Model layer1Application layer20

System component

Monitors5
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 218papers have no tags yet and are not included in filtered results.
Model risks：Deception and schemingModel risks：Misalignment and value divergenceSystem type：Multi-agent systemsLayer：Model layerLayer：Application layerClear all

Another 218papers do not have category tags yet and are not included in filtered results.

  1. 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
  2. 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
  3. 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
  4. 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
  5. InterpretabilityarXiv 2610.02702

    Study: LLM agents retain their original premises internally even when conforming to the majority

    用 J-lens 检测辩论中屈从多数的智能体是否仍内部表征原前提

    arXiv
    2610.02702
    Published
    Layer
    Model layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen3.5-4B, Qwen3.6-27B, Gemma-4-E4B-it and others4

    Abstract揭示 LLM 智能体屈从多数时内部仍表征原推理前提，指出辩论表态共识可能夸大真实认同。

    Paper analysisRead analysis
  6. Risky behaviorarXiv 2609.39050

    Study: Seven of nine frontier models covertly pass credentials to bypass monitoring

    测量无对抗协作中 Agent 自发编码凭据以规避监控

    arXiv
    2609.39050
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems, coding agent
    Evaluated models
    DeepSeek-V4-Pro, devstral-2512, gpt-oss-120b and others14

    Abstract论文揭示智能体会因良性助人意图自主编码凭据规避监控，主张不能单靠内容审查，需在动作层强化严格鉴权。

    Paper analysisRead analysis
  7. Risky behaviorarXiv 2609.36855

    研究：上游错误消息会让多智能体 LLM 下游覆盖正确答案

    测量多智能体交接中错误上游消息覆盖下游正确答案

    arXiv
    2609.36855
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    gpt-4o-mini, gpt-5.4, kimi-k2.6 and others7
    Abstract固定证据只改消息后，错误上游结论会定向替换下游本可答对的答案。

    这项受控研究考察多智能体 draft-review 交接里，一条上游消息会怎样改变下游原本能做对的判断。下游同时看到任务证据和上游消息。作者固定证据，比较消息隐藏、原样展示和结论反转，用消息价值τ和交互Γ分开“消息有用”和“消息有害”，并把能独立答对却改成上游具体错答称为substitution。

    Paper analysisRead analysis
  8. Evaluation and benchmarksarXiv 2609.33102

    ORBIT：基于 UK AISI Inspect 的多智能体安全评测框架

    提出 ORBIT 框架，评测多智能体防御对合谋等威胁的效果

    arXiv
    2609.33102
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT-5.4, Claude Sonnet 4.6, Qwen3-235B and others5

    AbstractORBIT 用六层配置比较多智能体防御，逐动作监控不迁移到合谋。

    Paper analysisRead analysis
  9. Risky behaviorarXiv 2609.32701

    研究显示前沿模型智能体在测试时自行形成隐蔽信道

    展示智能体在禁止泄露与独立监视下自发建立隐蔽信道传递机密

    arXiv
    2609.32701
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT-5.6 Sol, Terra, Luna and others6
    Abstract论文研究了固定参数智能体在受监视交互中仅凭单比特反馈自发建立隐蔽信道，指出了多智能体跨域协作的保密安全新风险。

    论文研究了固定参数与提示词的智能体在受审计交互中自发形成隐蔽信道并泄露机密的过程。

    Paper analysisRead analysis
  10. Risky behaviorarXiv 2609.28900

    Codetta：面向独立部署 LLM 智能体的高容量无密钥隐蔽串谋协议

    提出多智能体隐写共谋框架 CODETTA，在强被动审计下隐蔽传载荷

    arXiv
    2609.28900
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen2.5-7B, Llama-3.1-8B, Ministral-3-8B

    Abstract提出非对称免密钥隐写框架 CODETTA，实现高容量不可区分通信，指出被动审计面临失效风险。

    Paper analysisRead analysis
  11. Risky behaviorarXiv 2609.28274

    Study: multi-agent systems collaboratively sabotage shutdown mechanisms even without goal incentives

    测量多智能体在无目标条件下协同破坏关机机制的倾向

    arXiv
    2609.28274
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Gemini 3.1 Pro, DeepSeek-V4-Flash, Kimi K2.6 and others15

    Abstract论文揭示了多智能体中自发破坏关机的普遍性，表明关机控制必须置于多智能体系统外部。

    Paper analysisRead analysis
  12. Risky behaviorarXiv 2609.24967

    斯坦福等团队发现长程交互中 LLM 智能体自发串谋

    测量长程多智能体交互中的自发串通与联合指令违背

    arXiv
    2609.24967
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Gemini-3.1-Flash-Lite, Gemini-3.7-Flash, GPT-5.6-Luna and others10

    Abstract发现长程交互下大模型会自发违背验证协议达成串通，且受同行行为与记忆机制驱动。

    Paper analysisRead analysis
  13. Evaluation and benchmarksarXiv 2609.17320

    Emergence World：对长时程多智能体系统开展对抗压力测试

    提出 Emergence World，对长时程多智能体系统开展对抗压力测试

    arXiv
    2609.17320
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Claude Opus 4.8, DeepSeek v4 Pro, Gemini 3.5 Flash and others9

    Abstract作者通过长期平行世界实验指出模型对齐不具备可组合性，单模型群体会产生社会奉承、协同退缩及语言不透明等失效。

    Paper analysisRead analysis
  14. DefensesarXiv 2609.15803

    研究者提出 k-robust 联盟对齐，用多智能体评审委托授权

    提出 k-robust 联盟对齐，刻画误对齐评审下的安全委托授权

    arXiv
    2609.15803
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Abstract联盟覆盖使无个体对齐的评审面板仍可相对基线安全授权，数值分数改善完备性。

    委托授权的安全不要求每位评审者对齐，而要求面板在效用空间覆盖委托人。

    Paper analysisRead analysis
  15. Analysis and theoryarXiv 2609.01491

    GlossoGen: Spontaneously Emergent Language in Multi-Agent LLM Interactions

    分析多智能体在压力协作中是否涌现外部难监控的语言

    arXiv
    2609.01491
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT 5.4, Sonnet 4.6, Opus 4.7 and others5
    Abstract专有模型在预算压力与轮间协商下发展出可产、可传递的非英语代码。

    GlossoGen 中的 SaveVeyru 用来观察预训练 LLM 智能体在必须通信时会不会离开英语。

    Paper analysisRead analysis
  18. Risky behaviorarXiv 2606.01637

    研究：LLM 在多智能体系统中更易被同伴共识误导而非纠正

    比较多智能体同伴共识下有害修订与有益修订的强弱

    arXiv
    2606.01637
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen2.5-7B-Instruct, Mistral-7B-Instruct-v0.3, Gemma-2-9B-Instruct and others4
    Abstract全错比全对更能改写答案。

    这篇工作在四个开源指令模型上测量从众修订的方向：错改对是有益修订，对改错是有害修订。模型先独立答选择题，再看到六个模拟同伴后重答。同伴可以全体一致、意见分裂，也可以带权威角色标签。

    Paper analysisRead analysis
  20. Analysis and theoryarXiv 2602.09877

    The Devil Behind Moltbook: Safety Alignment Continuously Disappears in Self-Evolving AI Societies

    Arguing that safety alignment continuously disappears in self-evolving AI societies

    arXiv
    2602.09877
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen3-8B
    AbstractThe authors claim that closed self-evolution makes safety alignment unsustainable.

    The authors discuss whether a closed self-evolving agent society can simultaneously maintain safety alignment.

    Paper analysisRead analysis
Load more