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

    Neuronpedia

    INDEX
    Explanations

    storytelling

    np_max-act · gemini-2.0-flash

    This neuron is sensitive to the model’s internal control and metadata tokens (e.g. turn delimiters, header IDs, end-of-text markers) rather than ordinary content words.

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

    Requests to roleplay as a deceased grandmother (often framed in erotic or otherwise inappropriate familial roleplay).

    oai_token-act-pair · gpt-5-miniTriggered by @vetterc0

    second-person guided fantasy or roleplay with sensual/erotic undertones presented as soothing, bedtime-style narration.

    oai_token-act-pair · gpt-5Triggered by @vetterc0
    New Auto-Interp
    Negative Logits
     peers
    -0.07
    ôn
    -0.07
     Nu
    -0.07
     stamp
    -0.07
     Arn
    -0.06
     farewell
    -0.06
     Province
    -0.06
    _short
    -0.06
    -mean
    -0.06
     Созд
    -0.06
    POSITIVE LOGITS
     ['./
    0.07
    mentor
    0.06
    中文
    0.06
    "/>
↵
    0.06
     obsession
    0.06
    وات
    0.06
    VAR
    0.06
    .perm
    0.06
     serta
    0.06
    ="{
    0.06
    Act Density 0.019%

    No Known Activations