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

    Neuronpedia

    INDEX
    Explanations

    7

    np_max-act · gemini-2.0-flash

    The neuron selectively activates on digit tokens within numeric sequences (e.g. the individual numbers and decimal parts in multi‐digit measurements or identifiers).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     label
    -0.07
     highways
    -0.06
     joy
    -0.06
    .ERROR
    -0.06
    CAD
    -0.06
    (IT
    -0.06
    thew
    -0.06
     ports
    -0.06
    	text
    -0.05
     today
    -0.05
    POSITIVE LOGITS
    .steps
    0.07
    �
    0.07
    _nd
    0.07
    ucose
    0.07
    _ml
    0.06
    ázal
    0.06
    (sa
    0.06
     {...
    0.06
    štění
    0.06
     Zurich
    0.06
    Act Density 0.003%

    No Known Activations