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

    Neuronpedia

    INDEX
    Explanations

    variants

    np_max-act · gemini-2.0-flash

    This neuron detects speaker-label tokens (role prefixes) in a chat transcript, such as “GPT:”, “NAME_2:”, etc.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     독
    -0.06
    .Popup
    -0.06
    urry
    -0.06
    �
    -0.06
     inhal
    -0.06
    ENCH
    -0.06
    itemid
    -0.06
    itm
    -0.05
    survey
    -0.05
    unsafe
    -0.05
    POSITIVE LOGITS
     Ticaret
    0.07
     yytype
    0.07
    ()"↵
    0.07
     Initialize
    0.07
    APON
    0.07
     │
    0.06
    becue
    0.06
     yaklaşık
    0.06
    .contentSize
    0.06
    ##↵
    0.06
    Act Density 0.002%

    No Known Activations