© 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. 34249
    Prev
    Next
    INDEX
    Explanations

    code snippets and names

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

    The neuron strongly activates on programming‐language syntax tokens (identifiers, punctuation, and delimiters like braces and semicolons), effectively spotting code rather than natural text.

    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
    muñeca
    -1.05
    おかしい
    -1.02
     you
    -1.00
     mobiliario
    -0.98
    のではないか
    -0.96
     berichten
    -0.95
     compartida
    -0.94
    RETURNS
    -0.94
     czarne
    -0.94
    veless
    -0.93
    POSITIVE LOGITS
     Grille
    1.06
     €.
    0.98
     Smoothie
    0.94
     Pollo
    0.94
     ecco
    0.93
    避
    0.93
     Recife
    0.93
     Penthouse
    0.93
     Adhe
    0.92
     naves
    0.91
    Activations Density 0.004%

    No Known Activations