© 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. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    5. Gemma-2-9B
    7. Residual Stream - 16k
    9. 20-GEMMASCOPE-RES-16K
    11. 2177
    Prev
    Next
    INDEX
    Explanations

    the presence of significant statistical markers and concepts in scientific discussions

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

    complex legal and scientific terminology, particularly phrases involving statistical analysis or medical research methods.

    oai_token-act-pair · claude-3-5-haiku-20241022Triggered by @neilrathi

    paragraph-starting transitions or markers like "Finally," "However," "In," and "Thus" that introduce new sections or arguments.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi

    sentence-initial discourse markers and transitional phrases that signal logical flow, emphasis, or conclusions in formal/academic texts.

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

    sentence-initial words or phrases that begin new sentences or clauses in formal or academic text.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 20-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_20/width_16k/average_l0_68
    Prompts (Dashboard)
    24,576 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
     الحره
    -0.85
     ModelExpression
    -0.73
     فريبيس
    -0.71
    ſſung
    -0.63
     للاسماء
    -0.63
    :✨
    -0.63
     tartalomajánló
    -0.62
    iſen
    -0.61
    IsContent
    -0.59
    iſchen
    -0.58
    POSITIVE LOGITS
     imidlertid
    0.40
    Fortunately
    0.32
    Interestingly
    0.31
     således
    0.31
    Nevertheless
    0.31
     demikian
    0.30
     asimismo
    0.28
    Obviously
    0.28
     Fortunately
    0.28
     Enfin
    0.27
    Activations Density 1.235%

    No Known Activations