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

    Neuronpedia

    INDEX
    Explanations

    programming code

    np_max-act · gemini-2.0-flash

    The neuron activates on explanatory or commentary words and phrases in prose (e.g. “This,” “program,” “works,” “function”), so it’s effectively spotting narrative explanations rather than code.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
    .$$
    -0.07
     Tuesday
    -0.07
     Wednesday
    -0.06
     Cosmetic
    -0.06
    _ROOM
    -0.06
    ARGE
    -0.06
    时代
    -0.06
     Thursday
    -0.06
     wors
    -0.06
    Tank
    -0.06
    POSITIVE LOGITS
    _TRIANGLES
    0.07
     birlik
    0.06
    	UINT
    0.06
    Û
    0.06
    gresql
    0.06
     unitOfWork
    0.06
    )viewDidLoad
    0.06
    serviceName
    0.06
     ÜNİVERS
    0.06
    .vstack
    0.06
    Act Density 0.064%

    No Known Activations