© 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. 10-GEMMASCOPE-RES-131K
    7. 15437
    Prev
    Next
    INDEX
    Explanations

    Significance Statement

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

    This neuron fires on prominent title or headline words—that is, high-salience, often capitalized tokens that introduce sections, headings, or announcements.

    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
     Dinas
    -1.78
     horrifying
    -1.69
     自分
    -1.66
     emprego
    -1.62
     différentes
    -1.62
     pesso
    -1.59
     DIFFERENT
    -1.59
     첫
    -1.57
    
    -1.55
     ridiculous
    -1.55
    POSITIVE LOGITS
     may
    1.91
     would
    1.77
     should
    1.75
     might
    1.65
     did
    1.55
     you
    1.48
     noteworthy
    1.45
     unrelenting
    1.42
    Other
    1.41
     शुर
    1.41
    Activations Density 0.140%

    No Known Activations