© 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
    2. Gemma-2-27B
    3. 34-GEMMASCOPE-RES-131K
    4. 23794
    Prev
    Next
    INDEX
    Explanations

    place names and specific affiliations

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

    The neuron activates strongly on hyphen-joined place names (e.g. “Novara–Varallo”), so it’s detecting compound geographic names connected by a dash.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/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
    ↓
    -0.83
    WIN
    -0.82
    SNP
    -0.73
    😹
    -0.71
    賞
    -0.71
     Belize
    -0.71
     courage
    -0.70
    LETTER
    -0.70
     dare
    -0.69
    gorm
    -0.68
    POSITIVE LOGITS
     Huntsville
    0.93
     Danville
    0.91
     Lucknow
    0.88
     Peterborough
    0.86
     Carbide
    0.84
    Buffalo
    0.79
     Ithaca
    0.77
     나타
    0.77
     commutation
    0.77
    Tuk
    0.76
    Activations Density 0.037%

    No Known Activations