© 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. 10-GEMMASCOPE-RES-131K
    4. 102004
    Prev
    Next
    INDEX
    Explanations

    multi-lingual word endings and roots

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

    The neuron strongly activates on subword chunks that are common Latin/Greek roots or affixes (e.g. bio-, form-, te-, atome-, conceptual-) and thus identifies academic or scientific terminology.

    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
     a
    -1.93
     what
    -1.84
     of
    -1.82
     it
    -1.55
     vengono
    -1.52
     each
    -1.48
     who
    -1.44
     lessen
    -1.42
    amp
    -1.42
     to
    -1.42
    POSITIVE LOGITS
     этом
    2.02
    ==""){
    1.63
    cmx
    1.59
    %).
    1.55
     lakier
    1.53
     klienta
    1.52
     origins
    1.52
     podoba
    1.52
    
    1.49
    价比
    1.45
    Activations Density 0.055%

    No Known Activations