© 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. 13893
    Prev
    Next
    INDEX
    Explanations

    developerdescription text or code snippet

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

    This neuron fires on structural metadata cues in posts—words that mark headings or label fields (like “external,” “problem,” “url removed,” “COPYING,” “EMAIL PROTECTED”).

    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
     perplexing
    -1.84
     vehemently
    -1.82
     our
    -1.82
     meticulous
    -1.78
     superbly
    -1.76
     If
    -1.73
     When
    -1.71
     Creating
    -1.70
     Sometimes
    -1.69
     Often
    -1.66
    POSITIVE LOGITS
     Dinas
    1.75
     extremely
    1.69
    primarily
    1.68
    .='
    1.67
    mainly
    1.63
    clearly
    1.63
    ientras
    1.61
    arised
    1.60
     hà
    1.60
    generally
    1.59
    Activations Density 0.030%

    No Known Activations