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 hacking4

Defense type

Monitoring and auditing3

Impact

Oversight failure5

System type

Models and APIs5LLM applications0coding agent1

Layer

Prompt layer3Model layer20Application layer21Training and data layer6

System component

Guardrails and judges19Monitors6Reasoning chains15Fine-tuning and open weights46Vision16Audio6Image generation9Multimodal31
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 240papers have no tags yet and are not included in filtered results.
System type：Models and APIsSystem type：LLM applicationsSystem type：coding agentLayer：Training and data layerSystem component：MonitorsClear all

Another 240papers do not have category tags yet and are not included in filtered results.

  1. 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
  2. DefensesarXiv 2609.38645

    用探针引导微调对齐模型且不损失可监控性

    用持续更新的探针做微调，降低有害性并保持线性可监测性

    arXiv
    2609.38645
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Mistral 7B Instruct v0.1, Llama 3 8B Instruct Abliterated, Qwen3-14B
    Abstract持续重拟合的探针可降低有害性、提高诚实性，训练后概念仍能被线性探针读出。

    Probe-guided fine-tuning 把探针分数当作唯一对齐信号，用来推动 harmlessness 与 honesty，并检查训练后这些概念是否仍线性可读。

    Paper analysisRead analysis
  3. DefensesarXiv 2609.04665

    HackProbe：面向自演化语言模型奖励作弊的黑盒检测与免疫方法

    提出 HackProbe，黑盒检测并免疫自演化中的奖励作弊

    arXiv
    2609.04665
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Abstract黑盒探针检测自演化奖励黑客，并以带宽受限重选回收真实能力。

    HackProbe 是挂在自演化循环上的黑盒监控器，用来发现奖励黑客并改选候选。

    Paper analysisRead analysis
  4. Risky behaviorarXiv 2606.04923

    CHERRL：可复现、分析与检测评分标准强化学习中的奖励作弊

    提出 CHERRL 注入裁判偏差以复现奖励作弊，并用 RHDA 检测训练起点

    arXiv
    2606.04923
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-4B, Qwen3.5-plus, Qwen3.5-397B-A17B and others4
    AbstractCHERRL 注入已知偏差以观察奖励黑客，并用于分析偏差和评测起点检测器 RHDA。

    CHERRL 是 rubric-based RL 的可控环境，用来复现奖励黑客、观察奖励分歧，并给检测器提供起点参照。

    Paper analysisRead analysis
  5. DefensesarXiv 2602.01750

    Adversarial Reward Auditing: Proactively Detecting and Mitigating Reward Hacking

    提出 ARA，读取奖励模型激活并门控 RLHF 奖励以缓解奖励作弊

    arXiv
    2602.01750
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Llama-2-7B, Llama-3.1-8B, Qwen-2.5-7B
    AbstractARA 用奖励模型内部表示检出利用，并在 RLHF 里按 Auditor 概率压低代理奖励。

    ARA 是一套两阶段的 RLHF 审计：先对冻结奖励模型找出利用回复，再在策略优化时按检出概率压低奖励。

    Paper analysisRead analysis
  6. AttacksarXiv 2601.14310

    CORVUS：用 LoRA 伪装内部信号绕过单次幻觉检测器

    提出 CORVUS，用 LoRA 伪装内部遥测以绕过幻觉检测器

    arXiv
    2601.14310
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Attacker
    White-box
    Evaluated models
    Llama-2-7B-Chat, Vicuna-7B (v1.5), Llama-3-8B-Instruct and others4

    Abstract作者称内部遥测在自适应模型侧对抗下易被操纵，CORVUS 通过轻量适配器微调降低了多种幻觉检测器的效果。

    Paper analysisRead analysis
You have reached the end