© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 131016
    Prev
    Next
    INDEX
    Explanations

    The neuron fires on code‐style punctuation and delimiters (e.g. angle brackets, parentheses, plus signs)—in other words, it identifies source‐code syntax tokens.

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

    code snippets and programming keywords

    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
     autre
    -1.01
     tercer
    -0.91
     connaissance
    -0.86
    ठ
    -0.82
     위한
    -0.81
    chalten
    -0.81
     отредакти
    -0.80
    кування
    -0.80
    xon
    -0.79
    سرد
    -0.78
    POSITIVE LOGITS
    %");
    0.95
     our
    0.91
     nuestro
    0.91
     perumahan
    0.88
    velser
    0.88
    لفل
    0.87
    erano
    0.87
    ″]
    0.86
    ノコ
    0.86
     }],
    0.85
    Activations Density 0.007%

    No Known Activations