© 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. 34-GEMMASCOPE-RES-131K
    7. 64157
    Prev
    Next
    INDEX
    Explanations

    brand names, abbreviations, and symbols

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

    The neuron activates strongly on multi‐character tokens that include uppercase letters and/or digits (e.g. “CD8,” “CD4,” “YouTube,” “Supplex®”), indicating it’s detecting acronyms, brand names, and other alphanumeric sequences.

    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
     who
    -1.57
     this
    -1.56
     in
    -1.46
     truly
    -1.43
     what
    -1.34
     really
    -1.34
     that
    -1.32
     utterly
    -1.29
     all
    -1.27
     how
    -1.26
    POSITIVE LOGITS
     truff
    1.32
     kalender
    1.30
     lada
    1.27
     bombe
    1.27
     menyen
    1.26
     tirs
    1.26
     stund
    1.26
     legendar
    1.26
     yng
    1.26
     citroen
    1.24
    Activations Density 0.108%

    No Known Activations