INDEX
    Explanations

    language-specific words

    The neuron detects document-structuring tokens—headings, section breaks, and list/formatting markers that signal the outline or organization of the text.

    New Auto-Interp
    Negative Logits
     of
    0.28
     a
    0.25
     on
    0.24
     by
    0.24
     x
    0.24
     with
    0.23
     it
    0.23
     in
    0.22
     due
    0.22
     from
    0.22
    POSITIVE LOGITS
     hinzufügen
    0.26
    мпаваць
    0.26
    ясплат
    0.25
     ازيكم
    0.24
    पुस्तक
    0.24
     někol
    0.24
     தாவர
    0.24
    ിക്കാം
    0.23
    ocolate
    0.23
     ډاونلوډ
    0.23
    Act Density 0.611%

    No Known Activations