Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    INDEX
    Explanations

    Code

    np_max-act · gemini-2.0-flash

    This neuron reliably lights up on markup‐style tags (e.g. XML/HTML elements like `<exec>` or closing tags), i.e. it detects tokens that look like angle‐bracket tags.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     plan
    -0.08
     Introduction
    -0.07
    ько
    -0.07
     voices
    -0.07
    _fee
    -0.06
     worries
    -0.06
    身
    -0.06
     Ellis
    -0.06
     限
    -0.06
    .sh
    -0.06
    POSITIVE LOGITS
     sayısı
    0.07
    )))))↵
    0.06
     recibir
    0.06
     컴
    0.06
    ']].
    0.06
    是不
    0.06
     üzerinden
    0.06
    ))*(
    0.06
    */,↵
    0.06
     mutlaka
    0.06
    Act Density 0.035%

    No Known Activations