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

    Neuronpedia

    INDEX
    Explanations

    references to physical locations, specifically basements

    oai_token-act-pair · gpt-3.5-turbo

    references to basements and cellars

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Negative Logits
    yt
    -0.79
    uality
    -0.79
    nir
    -0.79
    itia
    -0.78
    Cola
    -0.74
    acting
    -0.74
    utan
    -0.73
    icer
    -0.73
    lda
    -0.72
    uncture
    -0.72
    POSITIVE LOGITS
     basement
    1.14
    stairs
    0.96
     hatch
    0.91
     lid
    0.88
     cellar
    0.88
     cavern
    0.83
     closet
    0.83
     Dwell
    0.83
     attic
    0.81
     crawl
    0.81
    Act Density 0.015%

    No Known Activations