INDEX
    Explanations

    The neuron fires on horoscope‐style metadata and astrology jargon—numbers/dates (e.g. years, days) and proper names of zodiac signs, planets, and aspects.

    New Auto-Interp
    Negative Logits
    Mag
    -0.07
     Ad
    -0.07
    -0.07
     měla
    -0.07
     náměstí
    -0.06
    ıyı
    -0.06
    时候
    -0.06
     само
    -0.06
    _fc
    -0.06
    лася
    -0.06
    POSITIVE LOGITS
     cigaret
    0.07
    coes
    0.06
    .xhtml
    0.06
    xic
    0.06
     Require
    0.06
    auss
    0.06
    brig
    0.06
    utures
    0.06
    know
    0.06
    แกรม
    0.06
    Act Density 0.004%

    No Known Activations