Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    INDEX
    Explanations

    attends to tokens that contain a numerical or letter-based identifier from graphical elements indexed with brackets

    oai_attention-head · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Head Attr Weights
    0:0.16
    1:0.13
    2:0.14
    3:0.05
    4:0.09
    5:0.23
    6:0.05
    7:0.10
    Negative Logits
    shub
    -0.33
     Allo
    -0.33
    uvo
    -0.30
    cue
    -0.29
     ledit
    -0.28
     通販
    -0.28
     allo
    -0.28
    yssey
    -0.28
    WEBPACK
    -0.28
     vál
    -0.27
    POSITIVE LOGITS
    UnusedPrivate
    0.43
    ReusableCell
    0.37
     himſelf
    0.36
    StoryboardSegue
    0.34
     myſelf
    0.34
    tagHelperRunner
    0.33
     themſelves
    0.33
    UnsafeEnabled
    0.33
     muſt
    0.32
     reaſon
    0.32
    Act Density 0.079%

    No Known Activations