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

    Neuronpedia

    INDEX
    Explanations

    Google API code

    np_max-act · gemini-2.0-flash

    The neuron activates on version‐style tokens (e.g. “v”, numeric release identifiers like “2018”, “202005”, etc.) in code package or filename paths, effectively spotting version numbers in namespace strings.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    New Auto-Interp
    Negative Logits
     DOT
    -0.07
     nid
    -0.07
     δο
    -0.06
     learnt
    -0.06
     بوده
    -0.06
    .PathVariable
    -0.06
     グ
    -0.06
     خواه
    -0.06
    继
    -0.06
     yön
    -0.06
    POSITIVE LOGITS
     concrete
    0.07
     Lab
    0.06
    사진
    0.06
     Card
    0.06
     Kiş
    0.06
    �
    0.06
     extra
    0.06
     M
    0.06
    .before
    0.06
    Github
    0.06
    Act Density 0.003%

    No Known Activations