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

    Neuronpedia

    INDEX
    Explanations

    code/programming

    np_max-act · gemini-2.0-flash

    The neuron activates on common function words—especially auxiliary and modal verbs (e.g. “that,” “been,” “will,” “be,” “don’t,” “think,” “im”)—rather than on content words or code tokens.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     cassette
    -0.07
     Goose
    -0.07
    _uniform
    -0.06
     Username
    -0.06
    -up
    -0.06
    �江
    -0.06
     přev
    -0.06
    _curve
    -0.06
    bsolute
    -0.06
     Khan
    -0.06
    POSITIVE LOGITS
    "]:↵
    0.06
     investigative
    0.06
    />
↵
    0.06
    InView
    0.06
     careless
    0.06
     اختل
    0.06
    ]+"
    0.05
    ObjectContext
    0.05
     differ
    0.05
    ));↵↵↵
    0.05
    Act Density 0.162%

    No Known Activations