© 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. 47728
    Prev
    Next
    INDEX
    Explanations

    signal frequency filtering

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

    This neuron responds to high-frequency function words and punctuation (common grammar markers), indicating it’s detecting the structural “glue” of text (stopwords like “the,” “or,” “for,” “such,” commas, etc.).

    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
    蜥
    -0.98
    orphism
    -0.83
     арт
    -0.82
    动静
    -0.81
    #+#
    -0.81
    スプレー
    -0.81
    chability
    -0.80
    MSI
    -0.79
    iov
    -0.77
     السياس
    -0.75
    POSITIVE LOGITS
     signal
    1.61
     Fourier
    1.48
     spectral
    1.41
     frequency
    1.40
     filter
    1.40
    Fourier
    1.39
     filters
    1.34
     Filter
    1.34
     spectrum
    1.34
     signals
    1.32
    Activations Density 0.093%

    No Known Activations