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

    Neuronpedia

    INDEX
    Explanations

    Period

    np_max-act · gemini-2.0-flash

    This neuron highlights descriptive “to‐do” or instruction phrases in code/problem statements—that is, it fires on words specifying the required actions (e.g. “read,” “lowercase,” “add spaces,” “write,” etc.).

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
    Avoid
    -0.07
     wsp
    -0.06
     yere
    -0.06
    poser
    -0.06
    _RT
    -0.06
    bull
    -0.06
    εβ
    -0.06
     egt
    -0.06
     Guy
    -0.06
     bw
    -0.06
    POSITIVE LOGITS
    Fantastic
    0.07
    Expanded
    0.06
    antically
    0.06
    abol
    0.06
    �
    0.06
    ">'.
    0.06
    ignment
    0.06
    _]
    0.06
     věc
    0.06
    ()))↵
    0.06
    Act Density 0.090%

    No Known Activations