© 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
    3. Gemma-2-27B
    5. 34-GEMMASCOPE-RES-131K
    7. 86943
    Prev
    Next
    INDEX
    Explanations

    specific words followed by context

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

    The neuron strongly activates on legal‐style citation elements—particularly court names and reporter volume/page numbers.

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

    formal document-structure and metadata markers (e.g., editorial tags, citation/license boilerplate, and institutional/legal section headers).

    oai_token-act-pair · gpt-5Triggered by @jyhe0408

    formatting elements in structured technical or legal documents, particularly colons and periods used as separators or delimiters.

    oai_token-act-pair · claude-4-5-sonnetTriggered 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
     salat
    -0.69
    不开
    -0.63
     pray
    -0.63
     Bakar
    -0.62
    abond
    -0.62
    라마
    -0.62
    нологи
    -0.61
    куп
    -0.60
    iras
    -0.60
    multer
    -0.60
    POSITIVE LOGITS
    الط
    0.72
    Senator
    0.69
    ೋ
    0.69
    phorus
    0.68
    ombus
    0.68
     independiente
    0.66
    idet
    0.66
    шено
    0.66
    ্ত
    0.65
    ପ
    0.65
    Activations Density 0.075%

    No Known Activations