© 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. 10-GEMMASCOPE-RES-131K
    7. 120013
    Prev
    Next
    INDEX
    Explanations

    programming keywords and identifiers

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

    underscores in variable names, particularly in code identifiers.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @bfgcn7frf5

    The neuron selectively activates on multi‐part identifier or class‐name tokens that contain internal separators (underscores or hyphens), e.g. fragmentary code or CSS/HTML class names composed of multiple words joined by “_” or “-.”

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/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
     of
    -2.03
    轰隆隆
    -1.66
     découver
    -1.64
    轰隆
    -1.62
    те
    -1.41
     kopen
    -1.39
    是非常
    -1.38
    d
    -1.38
    することも
    -1.38
    ar
    -1.37
    POSITIVE LOGITS
     zunächst
    1.68
    舠
    1.63
    
    1.63
    )--(
    1.54
     first
    1.51
     Aufenthalt
    1.51
    でしたが
    1.50
     how
    1.49
     colher
    1.47
    angerie
    1.46
    Activations Density 0.031%

    No Known Activations