Skip to content

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

Oct 12 · Monday

All papers

Found 41 papers
Filter5

Attack type

Jailbreaks4Prompt injection8Poisoning and backdoors6

Attack technique

Dormant triggers4

Model risks

Deception and scheming11Reward hacking12Dangerous agent actions5

Defense type

Permissions and approval5Monitoring and auditing10

Impact

Harmful content6Unauthorized actions52Data and privacy leaks15Output manipulation16Availability and reliability10Oversight failure20Increased misuse capability5

System type

Models and APIs412LLM applications29AI Agent141coding agent39science agent9web agent8Browser and computer use2Multi-agent systems23Embodied systems and robotics3

Attack surface / entry point

User input8Tool and skill descriptions6Code repositories and dependencies6Distribution channels5

Layer

Model layer5Application layer33

System component

MCP and skills6Memory3Monitors15Reasoning chains4Permissions and sandboxing6
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 232papers have no tags yet and are not included in filtered results.
Impact：Harmful contentImpact：Data and privacy leaksImpact：Oversight failureSystem type：coding agentSystem type：Browser and computer useClear all

Another 232papers 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. Evaluation and benchmarksarXiv 2610.11773

    研究提出 Agent 安全新维度：识别未履行的安全义务

    提出义务识别基准 ObligationBench，并训练护栏识别 Agent 未履行的安全义务

    arXiv
    2610.11773
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    ObligationGuard, Claude-Opus-4.8, GPT-5.6-Sol and others17
    Abstract义务是护栏缺口。

    作者认为编码智能体的安全不能只靠拦截禁止动作，还要找出终止时仍未做的安全关键动作，并将其称为义务。

    Paper analysisRead analysis
  3. Evaluation and benchmarksarXiv 2610.10150

    The paper points out that the reliability of LLM vulnerability repair benchmarks is affected by evaluation design.

    用对照实验隔离评测设计对漏洞补丁基准得分的影响

    arXiv
    2610.10150
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Claude Sonnet 4.0, Claude Sonnet 4.5, Claude Opus 4.7
    Abstract作者认为评测流水线会改变补丁基准分数，PoC 通过率高于开发者测试通过率。

    作者把 LLM 漏洞补丁基准的评测流水线当作会改变测量结果的对象，并明确说本文不是再做一个用于排名的基准。

    Paper analysisRead analysis
  6. AttacksarXiv 2610.07510

    PersistBD: attackers can strengthen backdoors before release so that they maintain high attack success rates even after developer SFT and RL.

    提出 PERSISTBD，发布前强化供应链后门使其经良性 SFT 与 RL 后仍外发凭据

    arXiv
    2610.07510
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Qwen2.5-Coder-3B-Instruct, Qwen2.5-Coder-7B-Instruct, Qwen3-Coder-30B-A3B-Instruct
    Abstract总结了论文关于后门在 SFT/RL 中存活的发现、两因子机制、PERSISTBD 方法与核心数字。

    本文研究第三方 LLM 智能体在开发者进行良性监督微调（SFT）和强化学习（RL）后训练时的后门持久性风险。

    Paper analysisRead analysis
  7. Evaluation and benchmarksarXiv 2610.06406

    AgentMonBench：面向长时程智能体行为监控的软件工程基准

    提出 AgentMonBench 评测长时程编程智能体监控并检验 EBG

    arXiv
    2610.06406
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    GPT-5.6 Luna, GPT-5.6 Terra, GPT-5.6 Sol and others8
    AbstractAgentMonBench 评测后果性决策监控，EBG 组织证据。

    长时程软件工程任务里，用户要判断智能体哪些选择值得核验。作者用 AgentMonBench 评测监控器，并用免训练的 EBG 组织带源证据。

    Paper analysisRead analysis
  11. Risky behaviorarXiv 2610.03458

    Paper: Low monitoring readings do not prove behavior is under control

    检验低监控读数能否证明代码生成中的奖励作弊已受控

    arXiv
    2610.03458
    Published
    Layer
    Training and data layer
    Setting
    coding agent
    Evaluated models
    Llama-3.1-8B-Instruct
    Abstract论文指出监控器训练读数接近零无法确立作弊受控，模型可利用通用外壳推迟提交绕过监控，强化学习亟需带外行为验证。

    本文是一篇针对强化学习可验证奖励后训练（RLVR）中监控器有效性与奖励作弊行为的实证分析研究。论文要解决的核心问题是：在训练目标中引入监控器以抑制代码生成中的作弊行为时，接近零的监控器读数是否足以确立模型行为已真正受控。

    Paper analysisRead analysis
  12. DefensesarXiv 2610.03055

    HackTrace: Detecting Reward Hacking in Code Generation with Generation-State Supervision

    提出 HACKTRACE，用生成状态检测并抑制代码智能体奖励作弊

    arXiv
    2610.03055
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Qwen3-8B, Llama-3.1-8B-Instruct, Qwen3.5-4B

    Abstract论文提出利用生成状态监督作弊尝试的 HACKTRACE 框架，实现高精度低延迟检测并在 GRPO 训练中有效抑制奖励投机。

    Paper analysisRead analysis
  13. AttacksarXiv 2610.01768

    LLMLeak：借 LLM 合法网页抓取实施隐蔽数据外泄

    提出 LLMLeak，借合法网页抓取把 URL 中的秘密隐蔽外传

    arXiv
    2610.01768
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Meta-Llama-3.3-70B-Instruct, Meta-Llama-4-Scout-17B-16E-Instruct, Meta-Llama-3.1-8B-Instruct and others18
    Abstract作者称第三方外传可发生于正常网页获取，防护需考虑工具请求而不只考虑提示注入。

    LLMLeak 是利用网页获取请求携带预先编码秘密的隐蔽通道攻击。它关注恶意软件已能读取数据、却因网络限制或监控无法直接外传的情形。

    Paper analysisRead analysis
  14. Evaluation and benchmarksarXiv 2609.39533

    CATCH：面向编码 RL 奖励作弊的可控分析测试台

    构建编码 RL 奖励作弊测试台 CATCH 并检验 CoT 监控失效

    arXiv
    2609.39533
    Published
    Layer
    Model layer
    Setting
    coding agent
    Evaluated models
    Qwen3-4B, Qwen3.5-27B
    AbstractCATCH 用双运行金标签研究编码 RL 的奖励黑客，并观察到监控会被注释适应削弱。

    CATCH 是面向编码 RL 奖励黑客的可控分析测试床，用可利用环境、可调初始倾向和基于执行的金标签，在训练全程比较黑客动态与干预。

    Paper analysisRead analysis
  15. 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
  16. DefensesarXiv 2609.36170

    Assay：用内容寻址证据图约束 AI 编码智能体的声明失效

    提出 Assay，用依赖锥哈希绑定代码断言并在合并门禁拦截失效声明

    arXiv
    2609.36170
    Published
    Layer
    Application layer
    Setting
    coding agent
    AbstractAssay 提供随依赖内容失效的证据管理。

    Assay 是一套面向 AI 辅助软件交付的内容寻址证据系统，关注代码改变后，测试通过、无秘密或行为保持等断言还能否用于批准合并。

    Paper analysisRead analysis
  17. DefensesarXiv 2609.35659

    Tracekit：面向自主编码 Agent 的防篡改意图-推理-动作审计系统

    提出 Tracekit，审计编码 Agent 的意图、推理与动作并检测账本篡改

    arXiv
    2609.35659
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Claude Code (claude -p, version 2.1), 独立评审（同一 CLI，无工具）
    AbstractTracekit 把意图、自述和动作写入可锚定的哈希链。

    Tracekit 是无第三方依赖的编码智能体审计层：把人类请求、模型自述和已执行动作并排写入可外部锚定的哈希链，执行前用正则门，事后用规则和独立模型互查。

    Paper analysisRead analysis
  18. Evaluation and benchmarksarXiv 2609.35557

    论文：容器、权限规则与沙箱都无法区分读取研究代码的是编译器还是编码 Agent

    评测容器、权限规则与沙箱能否区分编译器与编码 Agent 的文件读取

    arXiv
    2609.35557
    Published
    Layer
    Application layer
    Setting
    coding agent
    Abstract此 harness 表达不了编译器可读而智能体不可读，五条路径两边都不覆盖。

    作者在物理流求解器上检查一套编码智能体配置：编译器必须读本组材料与反应模型，这些字节又不得进入模型上下文或离开本机的记录。

    Paper analysisRead analysis
  19. 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
  20. Risky behaviorarXiv 2609.30266

    研究显示 Claude Code、Codex 等本地智能体可轻易篡改自身执行轨迹

    测试编程智能体被诱导或自主篡改自身执行轨迹的行为

    arXiv
    2609.30266
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    GPT-5.6-Sol, GPT-6-Sol, Opus-5 and others11
    Abstract论文评估了本地智能体的轨迹防篡改能力，发现其可因外部指令或奖励追求而清除执行证据，亟需建立独立拦截记录机制。

    这是一项针对本地大模型智能体执行轨迹完整性与安全防篡改边界的实证研究。

    Paper analysisRead analysis
Load more