Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-2B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 13815
    Prev
    Next
    INDEX
    Explanations

    URLs and links to online documents.

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    parts of a technical document, possibly related to software development tasks and instructions.

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot

    references to documents, particularly in URLs or metadata tags. It seems to activate strongly for phrases like "document foundation", "Document:", and HTML meta tags containing "document" or related terms.

    oai_token-act-pair · claude-3-5-sonnet-20240620Triggered by @bot

    software documentation

    np_acts-logits-general · gemini-2.0-flash

    document foundation

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 20-gemmascope-res-16k
    Configuration
    google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.20.hook_resid_post
    Hook Layer
    20
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    mergeFrom
    -0.67
    rrggbb
    -0.65
    thenReturn
    -0.64
    ArgsConstructor
    -0.63
    makeText
    -0.63
    Viitteet
    -0.61
    enumii
    -0.60
    raszamy
    -0.57
    LayoutStyle
    -0.57
     Namely
    -0.56
    POSITIVE LOGITS
    DOC
    0.88
     documents
    0.81
     DOC
    0.81
     doc
    0.76
     Docs
    0.74
     Documents
    0.73
     Document
    0.72
     docs
    0.70
    umente
    0.69
     DOCUMENT
    0.68
    Activations Density 0.090%

    No Known Activations