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

    Neuronpedia

    INDEX
    Explanations

    d

    np_max-act · gemini-2.0-flash

    The neuron strongly activates on numeric tokens and punctuation in metadata headers—e.g. version numbers, years, and parts of URLs (like “unity3d.com” or “2011”).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     factorial
    -0.07
     analytic
    -0.07
    (categories
    -0.06
    Boston
    -0.06
     manoe
    -0.06
     echoes
    -0.06
     Ion
    -0.06
    lemetry
    -0.06
     pin
    -0.06
     따
    -0.06
    POSITIVE LOGITS
    (Blueprint
    0.07
     totalmente
    0.07
    AMD
    0.07
    �
    0.07
    着
    0.07
    ||(
    0.07
    لعاب
    0.07
    -CN
    0.07
    动
    0.06
     ΔE
    0.06
    Act Density 0.000%

    No Known Activations