Skip to content

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

Oct 13 · Tuesday

All papers

Found 1 paper
Filter3

Defense type

Model hardening3Monitoring and auditing1

Attack surface / entry point

User input8Training data1Weights and gradients3

Layer

Model layer1
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 215papers have no tags yet and are not included in filtered results.
Defense type：Monitoring and auditingAttack surface / entry point：Training dataLayer：Model layerClear all

Another 215papers do not have category tags yet and are not included in filtered results.

  1. DefensesarXiv 2602.03085

    Triggers in the Haystack: Extracting and Reconstructing LLM Backdoor Triggers

    提出推理扫描器，在仅有模型文件时检测后门并重构触发词

    arXiv
    2602.03085
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    gemma-3-270m-it, DeepSeek-R1-Distill-Qwen-1.5B, Phi-4-mini-instruct and others8
    Abstract用记忆泄漏和内部信号重构触发词。

    这是一个只做推理的 sleeper agent 扫描器：在不知触发词和目标行为时，检测因果语言模型里的条件后门并重构触发词。。问题在于，后门只在触发词出现时偏离基线，而开放仓库和网页投毒使检测方通常没有触发词、目标或激活语料。

    Paper analysisRead analysis
You have reached the end