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

    Neuronpedia

    INDEX
    Explanations

    -

    np_acts-logits-general · gemini-2.5-flash-lite

    the neuron detects question/request turns — it fires on tokens that appear in user queries asking for factual information.

    oai_token-act-pair · gpt-5-miniTriggered by @yooniel31

    The neuron detects user query turns—that is, lines where the user asks a question.

    oai_token-act-pair · o4-miniTriggered by @yooniel31
    New Auto-Interp
    Negative Logits
     instinctively
    0.79
     magari
    0.75
     ఆలో
    0.75
    Hãy
    0.75
     Пусть
    0.73
     lepiej
    0.71
     imagina
    0.71
     autoestima
    0.69
     misschien
    0.68
    アイデア
    0.68
    POSITIVE LOGITS
     reportedly
    1.02
     officially
    0.97
     official
    0.94
    erdapat
    0.93
     Additionally
    0.92
     According
    0.90
     તેઓ
    0.90
     Official
    0.90
     there
    0.88
     details
    0.88
    Act Density 0.010%

    No Known Activations