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

    Neuronpedia

    INDEX
    Explanations

    the

    np_max-act · gemini-2.0-flash

    the neuron responds to repeated or boilerplate text—tokens that appear many times in a duplicated/templated phrase or repeatedly reiterated sentence fragments.

    oai_token-act-pair · gpt-5-miniTriggered by @yooniel31
    New Auto-Interp
    Negative Logits
     Characteristics
    -0.08
     shelves
    -0.07
     offences
    -0.07
     Roger
    -0.07
     Guard
    -0.06
     shelf
    -0.06
     Ты
    -0.06
     northern
    -0.06
     approves
    -0.06
     redevelopment
    -0.06
    POSITIVE LOGITS
    ะแ
    0.06
    _TYPED
    0.06
    .man
    0.06
    roleum
    0.06
     zaměř
    0.06
     RaycastHit
    0.06
     Specifies
    0.06
     dní
    0.06
    scious
    0.06
     Στο
    0.06
    Act Density 0.023%

    No Known Activations