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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 25828
    Prev
    Next
    INDEX
    Explanations

    foreign punctuation and words

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

    The neuron reliably fires on code-related tokens—especially programming keywords, identifiers, protocol/class names, and punctuation elements—i.e. pieces of source‐code syntax.

    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
     on
    -2.00
     only
    -1.73
     first
    -1.64
     all
    -1.63
     just
    -1.63
     started
    -1.61
     even
    -1.61
     most
    -1.55
     mostly
    -1.51
     for
    -1.50
    POSITIVE LOGITS
     trám
    1.59
    を買う
    1.48
    際に
    1.47
     serpiente
    1.46
     "."
    1.42
    €“
    1.41
     Dinas
    1.41
    eload
    1.40
     seleccionamos
    1.40
     sám
    1.38
    Activations Density 0.097%

    No Known Activations