INDEX
    Explanations

    Conversational/extracted text

    This neuron is activated by the special header‐and‐separator tokens that mark the start, end, and boundaries of metadata in the document (e.g. “<|start_header_id|>”, “<|end_header_id|>”, “<|begin_of_text|>”, etc.).

    New Auto-Interp
    Negative Logits
    ?>"
    -0.07
     instancia
    -0.06
     kullanılır
    -0.06
    	old
    -0.06
    lot
    -0.06
    ging
    -0.06
    	address
    -0.06
    Các
    -0.06
    -0.06
    	unit
    -0.06
    POSITIVE LOGITS
    toISOString
    0.06
    역시
    0.06
     مبانی
    0.06
     originated
    0.06
    igli
    0.06
     DBNull
    0.06
    0.06
     seis
    0.06
     Comb
    0.06
    δόν
    0.06
    Act Density 0.009%

    No Known Activations