INDEX
    Explanations

    This neuron activates on the phrase “on behalf of,” flagging expressions of representation or sponsorship.

    New Auto-Interp
    Negative Logits
     Taş
    -0.06
    ارش
    -0.06
     гориз
    -0.06
    urger
    -0.06
    ایط
    -0.06
    /i
    -0.06
     IDisposable
    -0.06
     gắn
    -0.06
    -Dec
    -0.06
    "There
    -0.06
    POSITIVE LOGITS
     behalf
    0.15
    .vo
    0.07
     ego
    0.07
    ля
    0.07
     heav
    0.07
    Campaign
    0.07
    To
    0.07
    0.07
    0.07
     representations
    0.07
    Act Density 0.006%

    No Known Activations