© 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. 128346
    Prev
    Next
    INDEX
    Explanations

    This neuron detects spans of natural‐language prose—typically descriptive comments or documentation sentences—embedded in code.

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

    number minus

    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
     innumerable
    -1.51
     nine
    -1.49
    -
    -1.48
     eight
    -1.46
     seven
    -1.44
     five
    -1.41
     Apparently
    -1.41
     first
    -1.38
     because
    -1.36
     Not
    -1.36
    POSITIVE LOGITS
    ,
    2.05
    "+
    1.92
     '',
    1.57
     '
    1.52
    \
    1.52
     eſt
    1.45
     $\&$
    1.45
    ',
    1.42
    ,\\
    1.41
    verhalten
    1.41
    Activations Density 0.398%

    No Known Activations