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

    Neuronpedia

    INDEX
    Explanations

    incorrect, false, wrong

    np_max-act · gemini-2.0-flash

    The neuron detects tokens that signal a correction or negation of a preceding statement (e.g. words like “incorrect,” “isn’t,” “false,” “wrong,” “quite right,” etc.).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
    ंब
    -0.06
     outfit
    -0.06
    Если
    -0.06
    IntoConstraints
    -0.06
    Utf
    -0.06
    امبر
    -0.06
    ソ
    -0.06
    ฺ
    -0.05
    Sup
    -0.05
    aklı
    -0.05
    POSITIVE LOGITS
     induces
    0.08
     inexperienced
    0.07
    _qs
    0.07
    (int
    0.07
     Shea
    0.07
     aster
    0.07
     Url
    0.07
    (CH
    0.06
    HAL
    0.06
     переж
    0.06
    Act Density 0.044%

    No Known Activations