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

    Neuronpedia

    INDEX
    Explanations

    phrases related to avoiding potential negative outcomes

    oai_token-act-pair · gpt-3.5-turbo

    phrases related to avoiding negative outcomes or risks

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Negative Logits
    soDeliveryDate
    -0.81
     largeDownload
    -0.77
    REL
    -0.71
    çľ
    -0.65
     Shap
    -0.65
    é¾įåĸļ士
    -0.65
     Compass
    -0.63
    å¦
    -0.62
     Frontier
    -0.61
    ãģ®é
    -0.60
    POSITIVE LOGITS
     altogether
    1.07
     pitfalls
    1.06
     harm
    0.94
     costly
    0.84
     misuse
    0.83
     temptation
    0.82
     harmful
    0.82
     bloodshed
    0.82
     accidental
    0.81
     disruption
    0.81
    Act Density 0.339%

    No Known Activations