© 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. 68030
    Prev
    Next
    INDEX
    Explanations

    code keywords

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

    The neuron strongly activates on tokens that start new statements or sections—especially uppercase SQL keywords like “PRIMARY” or other words appearing as the first (capitalized) token of a sentence or code block.

    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
     also
    -1.07
    ～～
    -1.03
    .]
    -0.97
    ov
    -0.95
     integrand
    -0.95
     even
    -0.95
    il
    -0.94
    BIOGRAPHY
    -0.93
     Ausdrücke
    -0.93
     And
    -0.92
    POSITIVE LOGITS
     célè
    1.32
     éte
    1.28
     surtido
    1.26
     muñeco
    1.23
     čer
    1.23
     bisous
    1.20
    あー
    1.20
     éta
    1.17
     obé
    1.17
     améli
    1.16
    Activations Density 0.098%

    No Known Activations