INDEX
    Explanations

    Random text

    Neuron 4 responds to capitalized or title-case tokens—essentially marking proper names, places, acronyms, and other named-entity mentions.

    New Auto-Interp
    Negative Logits
     превыш
    -0.07
     الرح
    -0.07
     تط
    -0.07
     antique
    -0.07
     tert
    -0.06
    انية
    -0.06
    ्रय
    -0.06
    ุทร
    -0.06
    .pretty
    -0.06
    ційного
    -0.06
    POSITIVE LOGITS
    _FIFO
    0.06
    ******/↵↵
    0.06
     irony
    0.06
    claims
    0.06
     گل
    0.06
    UIFont
    0.06
    )),↵
    0.05
    tie
    0.05
     Claim
    0.05
    SOR
    0.05
    Act Density 0.539%

    No Known Activations