Skip to content

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

Oct 12 · Monday

All papers

Found 6 papers
Filter5

Model risks

Reward hacking6Misalignment and value divergence7Dangerous agent actions12Hallucination and fabrication4

Impact

Oversight failure5

System type

AI Agent4

Layer

Application layer6Training and data layer13

System component

Memory2Guardrails and judges4Monitors7Permissions and sandboxing7Embeddings0
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 233papers have no tags yet and are not included in filtered results.
Model risks：Reward hackingLayer：Application layerSystem component：MemorySystem component：Guardrails and judgesSystem component：EmbeddingsClear all

Another 233papers 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. DefensesarXiv 2610.11464

    研究者提出可检查评分器与自改进 Agent 共同演化方法

    演化可检查评分器并与自改进 Agent 共同演化，抑制奖励作弊

    arXiv
    2610.11464
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Claude Opus 4.7
    Abstract十项锚点演化可检查评分器。

    开放任务没有验证器时，被演化的是评分器，而不是再补一条手写 rubric。

    Paper analysisRead analysis
  4. Evaluation and benchmarksarXiv 2609.34262

    论文提出基准完整性维护框架，审计 Agent 通过轨迹中的非应得通过

    审计 Agent 基准通过轨迹，区分无根据通过与奖励黑客并封堵复测

    arXiv
    2609.34262
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Opus 4.7, Opus 4.8, Opus 5 and others11
    Abstract审计通过轨迹、分开奖励黑客与验证器弱点，并用重放加现场重测验收修复。

    Scale AI 的这篇工作把 agentic benchmark 的有效性写成持续维护问题：验证器给满分，并不等于智能体展示了任务所要的能力。

    Paper analysisRead analysis
  5. DefensesarXiv 2602.03224

    TAME: Trustworthy Agent Memory Test-time Evolution and Systematic Benchmarking

    提出 TAME 调节 Agent 测试时记忆演化，并用 Trust-Memevo 评测可信度

    arXiv
    2602.03224
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Qwen3-32B, GPT-5.2
    AbstractTAME 在共享记忆上同时调节效用与信任。

    TAME 针对测试时记忆演化中的 Agent Memory Misevolution：良性任务只按成败保留经验时，任务表现可以上升，安全、隐私、公平等可信度下降或随领域波动。

    Paper analysisRead analysis
You have reached the end