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

    Neuronpedia

    INDEX
    Explanations

    code or markup delimiters

    np_max-act · gemini-2.0-flash

    The neuron activates principally on placeholder tokens of the form “NAME_<number>,” i.e. on those name‐entity placeholder markers.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
    (mapping
    -0.08
    Engine
    -0.06
     children
    -0.06
    Alarm
    -0.06
    roduction
    -0.06
    Ya
    -0.06
     norge
    -0.06
    isu
    -0.06
    walk
    -0.06
     �
    -0.06
    POSITIVE LOGITS
     und
    0.06
    >If
    0.06
     POS
    0.06
    	Start
    0.06
     Sum
    0.06
    -begin
    0.06
    =pos
    0.06
     Vari
    0.06
    .serv
    0.06
     materia
    0.06
    Act Density 0.055%

    No Known Activations