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

    Neuronpedia

    INDEX
    Explanations

    spanish/portuguese "can"

    np_max-act · gemini-2.0-flash

    The neuron strongly activates on conjugations of the verb “poder” (e.g. “pode,” “puede,” “puedes,” “possam”), i.e. modal words meaning “can/may” in Portuguese and Spanish.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8

    phrases related to health and safety regulations.

    oai_token-act-pair · gpt-4o-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     that
    -0.09
     on
    -0.07
     as
    -0.07
    18
    -0.06
     scholarship
    -0.06
     On
    -0.06
    3
    -0.06
     Rally
    -0.06
    YGON
    -0.06
     azi
    -0.06
    POSITIVE LOGITS
     può
    0.10
     could
    0.10
     может
    0.10
     may
    0.09
     can
    0.09
     μπορεί
    0.09
     possono
    0.09
     могут
    0.09
     можуть
    0.08
     можете
    0.08
    Act Density 0.075%

    No Known Activations