Skip to content

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

Oct 12 · Monday

All papers

Found 3 papers
Filter2

Model risks

Misalignment and value divergence3Dangerous agent actions3

System type

AI Agent3

Layer

Application layer3

System component

Memory3
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 261papers have no tags yet and are not included in filtered results.
Model risks：Misalignment and value divergenceSystem component：MemoryClear all

Another 261papers do not have category tags yet and are not included in filtered results.

  1. Evaluation and benchmarksarXiv 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
  2. 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
  3. Evaluation and benchmarksarXiv 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
You have reached the end