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

    Neuronpedia

    INDEX
    Explanations

    conjunctions followed by a contrast or opposing idea

    oai_token-act-pair · gpt-3.5-turbo

    instances of the word "but"

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Negative Logits
    onte
    -0.72
     Became
    -0.64
    obyl
    -0.62
    \":
    -0.60
    çĶ°
    -0.60
    èĢħ
    -0.59
     Entered
    -0.59
    ãģ®
    -0.58
    encia
    -0.58
    pat
    -0.58
    POSITIVE LOGITS
     nevertheless
    1.18
     nonetheless
    1.07
     suffice
    0.91
     hey
    0.91
    rarily
    0.78
     surely
    0.74
     insofar
    0.73
    orically
    0.72
     undeniably
    0.71
     alas
    0.70
    Act Density 0.202%

    No Known Activations