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

    Neuronpedia

    INDEX
    Explanations

    upward

    np_max-act · gemini-2.0-flash

    the neuron detects factual/reporting content—tokens that signal concrete facts, dates, numbers, or other informationally salient words in news-like passages.

    oai_token-act-pair · gpt-5-miniTriggered by @vetterc0
    New Auto-Interp
    Negative Logits
    directive
    -0.07
     Knot
    -0.07
    Yes
    -0.07
    .rm
    -0.07
     Fransız
    -0.06
    _cs
    -0.06
     qp
    -0.06
    衛
    -0.06
     Twe
    -0.06
     ErrorCode
    -0.06
    POSITIVE LOGITS
    �
    0.06
     BOOST
    0.06
    __),
    0.06
     taking
    0.06
    .Services
    0.06
     request
    0.06
    ‌
    0.06
    -pass
    0.06
     Taking
    0.06
    พร
    0.06
    Act Density 0.077%

    No Known Activations