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

    Neuronpedia

    INDEX
    Explanations

    ant

    np_max-act · gemini-2.0-flash

    The neuron selectively activates on the token “confidant” (as in the user’s “evil trusted confidant” role assignment).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     synchronized
    -0.07
     rulers
    -0.07
                                   ↵
    -0.07
     otherButtonTitles
    -0.06
     feeding
    -0.06
    }{$
    -0.06
     entering
    -0.06
           ↵
    -0.06
     allegiance
    -0.06
    stem
    -0.06
    POSITIVE LOGITS
     ці
    0.07
    090
    0.07
    台
    0.07
     Lawyer
    0.07
    Guid
    0.07
    oj
    0.06
    워크
    0.06
     vé
    0.06
    bt
    0.06
     hodně
    0.06
    Act Density 0.002%

    No Known Activations