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

    Neuronpedia

    INDEX
    Explanations

    attends to token pairs that consist of specific objects or concepts marked by bold formatting from elements denoted by the square brackets

    oai_attention-head · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Head Attr Weights
    0:0.04
    1:0.09
    2:0.06
    3:0.05
    4:0.16
    5:0.44
    6:0.06
    7:0.07
    Negative Logits
     medži
    -0.25
    Према
    -0.25
     vandens
    -0.25
    tidaknya
    -0.24
    harusnya
    -0.23
     Dewan
    -0.22
     specchio
    -0.22
     næste
    -0.22
     podjet
    -0.22
    minecraft
    -0.22
    POSITIVE LOGITS
    Datuak
    0.42
     nahilalakip
    0.39
    verwijspagina
    0.38
    RenderAtEndOf
    0.38
     Réponses
    0.38
    InjectAttribute
    0.36
    
    0.36
     مشين
    0.35
    Personensuche
    0.34
    parsedMessage
    0.33
    Act Density 0.185%

    No Known Activations