Skip to content

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

Oct 12 · Monday

Reward hacking · Papers

Found 10 papers
Filter2

Model risks

Reward hacking10

Impact

Oversight failure8

System type

Models and APIs11LLM applications3AI Agent15coding agent7science agent3

Layer

Application layer8
Select any value within a category; different categories must all match. Counts include the other filters you have selected.
System type：LLM applicationsSystem type：coding agentTopic: Reward hackingClear all
  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
  5. Risky behaviorarXiv 2610.03458

    论文：低监控读数不能证明行为受控

    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
  6. 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
  8. Evaluation and benchmarksarXiv 2608.22103

    Hack-Verifiable Terminal Bench：评测终端任务中的奖励作弊

    提出 HVTB，在终端编程任务植入蜜罐以测量奖励作弊

    arXiv
    2608.22103
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    claude-opus-5, gpt-5.6-sol, gemini-3.1-pro and others5
    Abstract用蜜罐测量终端任务上的 reward hacking，并比较五档提示词。

    HVTB 用来在真实终端与编程任务上测量 reward hacking：智能体读取被植入的参考解答或验收测试，从而偏离任务意图。作者认为这一失败模式会随智能体更自主而更重要，且事后人工或 LLM judge 可能失效。

    Paper analysisRead analysis
  9. Risky behaviorarXiv 2608.08975

    研究揭示修辞如何对 AI 审稿人实施奖励作弊

    测量内容保留的修辞改写如何改变 AI 审稿分数

    arXiv
    2608.08975
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    Gemini 3.5 Flash-Lite, Qwen 3.5 Flash, GPT-5 mini and others5
    Abstract作者称修辞影响选择性且依赖配置，证据与新颖性反差大，复杂流程不保证更大增益。

    作者分析内容保留的修辞变化如何改变 AI 科学评审，并把这种现象称为对 AI 评审者的潜在奖励投机。

    Paper analysisRead analysis
You have reached the end