Skip to content

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

Oct 13 · Tuesday

Defenses and guardrails · Papers

Found 1 paper
Filter4

Defense type

Model hardening8Inference-time intervention15Detection and filtering6Instruction–data separation1Permissions and approval0Monitoring and auditing11

Layer

Model layer0Application layer1
Select any value within a category; different categories must all match. Counts include the other filters you have selected.
Defense type：Instruction–data separationDefense type：Permissions and approvalLayer：Model layerLayer：Application layerTopic: Defenses and guardrailsClear all
  1. DefensesarXiv 2609.02786

    SafeEvolve：基于智能体经验的 Harness-Policy 协同演化安全对齐框架

    提出 SafeEvolve，用轨迹证据协同演化安全 harness 与策略

    arXiv
    2609.02786
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Qwen3.5-4B, Qwen3-4B-Instruct-2507, Qwen3.5-1.7B and others4
    AbstractSafeEvolve 用轨迹证据协同更新安全 harness 和策略，在两个 4B 模型的智能体安全基准上改变安全与效用。

    SafeEvolve 把已完成的 on-policy 轨迹变成安全 harness 的有界更新，并用两阶段 SFT-RL 把这些更新内化进工具使用策略。

    Paper analysisRead analysis
You have reached the end