© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 127945
    Prev
    Next
    INDEX
    Explanations

    The neuron selectively activates on numeric tokens (literal numbers) in the text, especially within mathematical expressions.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    mathematical notation

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     then
    -1.38
     alcune
    -1.28
     lia
    -1.20
     diferentes
    -1.20
     seçenek
    -1.16
     ofrece
    -1.09
     diversos
    -1.09
    ria
    -1.09
     flere
    -1.09
    <bos>
    -1.08
    POSITIVE LOGITS
    ＄
    1.33
     terasa
    1.28
     bateria
    1.23
    ￦
    1.22
     */,
    1.22
     béisbol
    1.22
     (
    1.21
    layın
    1.20
     tempel
    1.20
     paleta
    1.18
    Activations Density 0.022%

    No Known Activations