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

    Neuronpedia

    INDEX
    Explanations

    ens

    np_max-act · gemini-2.0-flash

    The neuron strongly activates on capitalized tokens and subword pieces of proper nouns or acronyms—that is, it’s a “named‐entity” detector.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8

    discussions about philosophical paradoxes related to motion and position.

    oai_token-act-pair · gpt-4o-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
    -ton
    -0.06
    SmartyHeaderCode
    -0.06
    bstract
    -0.06
     {}));↵
    -0.06
    .OnClickListener
    -0.06
     Tie
    -0.06
    Td
    -0.06
    .AWS
    -0.06
    formatter
    -0.06
     川
    -0.06
    POSITIVE LOGITS
     legs
    0.07
     incorpor
    0.07
     çevres
    0.07
     많이
    0.06
     apenas
    0.06
     adjustments
    0.06
     hodně
    0.06
     nop
    0.06
    805
    0.06
    ekte
    0.06
    Act Density 0.102%

    No Known Activations