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

    Neuronpedia

    INDEX
    Explanations

    negative sentiment

    np_max-act · gemini-2.0-flash

    The neuron strongly activates on anonymized name‐placeholders (tokens like “NAME_1,” “NAME_2,” etc.), i.e. it detects those dummy name tokens.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
    arus
    -0.07
     threw
    -0.07
     حو
    -0.07
     JW
    -0.07
     beck
    -0.06
     holes
    -0.06
    ,result
    -0.06
    ΣΤ
    -0.06
    xford
    -0.06
    	data
    -0.06
    POSITIVE LOGITS
    TagName
    0.07
     Tib
    0.07
    ／:
    0.06
    Rename
    0.06
    	EIF
    0.06
    ivr
    0.06
    _buff
    0.06
    clientId
    0.06
    .testng
    0.06
    _published
    0.06
    Act Density 0.042%

    No Known Activations