© 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. 22-GEMMASCOPE-RES-131K
    7. 128840
    Prev
    Next
    INDEX
    Explanations

    The neuron is tuned to pick out very common function words and punctuation (i.e. high‐frequency “stop” tokens).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    "i.e." explanations

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/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
     comarca
    -0.90
     lavora
    -0.85
    odyne
    -0.85
    itados
    -0.85
    Conseils
    -0.84
     slew
    -0.81
     Undang
    -0.81
    ORAGE
    -0.79
    あら
    -0.79
    Piotr
    -0.79
    POSITIVE LOGITS
     "
    1.84
     “
    1.44
     cioè
    1.25
    也就是
    1.24
     '
    1.16
    いわゆる
    1.16
     sogenannte
    1.14
     például
    1.03
     sogenannten
    1.02
    ":
    1.02
    Activations Density 0.068%

    No Known Activations