© 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. 128023
    Prev
    Next
    INDEX
    Explanations

    The neuron selectively fires on numeric and number‐related tokens (digits, colons, hyphens) used in citations, time stamps, page/line numbers, and similar formatted references.

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

    specifying page, line, or nucleotide positions

    np_acts-logits-general · gemini-2.5-flash-lite

    page and line number references in documents.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi
    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
     two
    -1.09
     with
    -1.01
     several
    -0.94
     was
    -0.91
     three
    -0.91
     while
    -0.90
     Why
    -0.81
     mildly
    -0.81
     me
    -0.77
     time
    -0.77
    POSITIVE LOGITS
    //*[@
    1.00
     části
    0.99
    starting
    0.98
     starting
    0.92
    beginning
    0.91
    的笑
    0.90
     sixth
    0.90
     begint
    0.89
     Artículo
    0.89
     problèmes
    0.88
    Activations Density 0.081%

    No Known Activations