© 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. 80023
    Prev
    Next
    INDEX
    Explanations

    branching, roots, reference, IDE

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

    The neuron consistently activates on tokens that contain numeric characters (e.g. digits, version numbers, timestamps), so it’s detecting number‐containing tokens.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    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
     面包
    -0.93
    ajuku
    -0.91
    Boxscore
    -0.87
     Kampagne
    -0.87
     课
    -0.86
    bcd
    -0.85
     উপ
    -0.84
     Aufführung
    -0.83
     계획
    -0.82
    Métodos
    -0.81
    POSITIVE LOGITS
     others
    0.98
    others
    0.87
    Others
    0.87
    ikti
    0.87
    empty
    0.86
     reine
    0.85
     empty
    0.84
    पर
    0.80
     فقط
    0.79
     zaj
    0.79
    Activations Density 0.087%

    No Known Activations