INDEX
    Explanations

    Type/Method

    This neuron responds to Portuguese terms that introduce categories—especially the word “tipo” (type) when listing different kinds of servers.

    New Auto-Interp
    Negative Logits
     erected
    -0.07
     enviado
    -0.06
    ieg
    -0.06
    yscale
    -0.06
    liche
    -0.06
     coated
    -0.06
    -0.06
    自己
    -0.06
     naveg
    -0.05
    ']));
    -0.05
    POSITIVE LOGITS
     Lana
    0.07
    /edit
    0.07
    Statistic
    0.07
    рь
    0.06
     ICT
    0.06
    ина
    0.06
     whites
    0.06
    ign
    0.06
    locking
    0.06
     authoritative
    0.06
    Act Density 0.059%

    No Known Activations