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

    Neuronpedia

    INDEX
    Explanations

    Leaked texts

    np_max-act · gemini-2.0-flash

    specific vocabulary related to military, warfare, and strategic concepts.

    oai_token-act-pair · gpt-4oTriggered by @tcai

    This neuron responds to language evoking secrets or clandestine revelations—words like “purloined,” “leaks,” “rumours,” “occult,” and similar terms signaling hidden or covert information.

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

    in-universe lore or flavor-text passages, especially those describing secret/rumored artifacts or leaked manuals with backstory and mystique.

    oai_token-act-pair · gpt-5Triggered by @vetterc0
    New Auto-Interp
    Negative Logits
     سیستم
    -0.07
    .texture
    -0.07
     τελ
    -0.06
    Newton
    -0.06
    亚
    -0.06
    -performance
    -0.06
     Hamp
    -0.06
    TTY
    -0.06
     occurrences
    -0.06
     Trent
    -0.06
    POSITIVE LOGITS
    exampleInputEmail
    0.07
     salvar
    0.07
    ...")↵
    0.07
    )↵↵↵↵
    0.07
     tslib
    0.07
    [image
    0.06
     "\
    0.06
     nomin
    0.06
     полит
    0.06
     пис
    0.06
    Act Density 0.326%

    No Known Activations