© 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. 128273
    Prev
    Next
    INDEX
    Explanations

    This neuron fires on programming‐language syntax tokens—e.g. keywords (var/def/import), punctuation (parentheses, semicolons, braces), operators (//, =), and other code markers—i.e. it detects code snippets.

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

    code structure and methods

    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
     exag
    -0.85
     desenho
    -0.84
     fright
    -0.84
    рги
    -0.84
     jist
    -0.82
     between
    -0.80
     edifício
    -0.80
    rsen
    -0.79
    需求
    -0.79
     morgue
    -0.79
    POSITIVE LOGITS
     method
    1.38
     using
    1.30
     methods
    1.15
     utilisant
    1.14
    Method
    1.12
     méthode
    1.09
    另一种
    1.08
     alternatives
    1.06
     alternative
    1.02
     pomocí
    1.02
    Activations Density 0.046%

    No Known Activations