INDEX
    Explanations
    New Auto-Interp
    Negative Logits
     bottlenecks
    0.40
     components
    0.38
     stencil
    0.37
     modules
    0.37
     topography
    0.36
     tableaux
    0.36
     functional
    0.36
     implement
    0.36
     substituents
    0.36
     outset
    0.35
    POSITIVE LOGITS
    yes
    0.51
    sorry
    0.51
     হ্যাঁ
    0.49
    Yes
    0.48
    thank
    0.47
    https
    0.47
    Sorry
    0.46
    yeah
    0.45
    Спасибо
    0.44
    indeed
    0.44
    Act Density 1.927%

    No Known Activations