© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-9B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 16218
    Prev
    Next
    INDEX
    Explanations

    technical terms and concepts related to digital computers and their functions

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

    formatting elements in technical or educational text, especially paragraph breaks, bullet points, and year dates.

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

    common function words and punctuation that signal sentence or list structure (e.g., copular/linking markers, structural punctuation, years/hyphenation cues).

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

    definite articles and conjunctions in formal, encyclopedic text describing technical or historical topics.

    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
    featureID
    -0.86
    setupUi
    -0.74
     فريبيس
    -0.73
    IVEREF
    -0.69
    最快更新
    -0.68
    twimg
    -0.66
     nakalista
    -0.66
    ConstraintMaker
    -0.65
    StoryboardSegue
    -0.65
    WireFormatLite
    -0.65
    POSITIVE LOGITS
     concepts
    0.55
     commonly
    0.54
     importance
    0.54
     technology
    0.53
     primarily
    0.52
     examples
    0.49
     definition
    0.49
     technologies
    0.48
     functions
    0.47
     concept
    0.47
    Activations Density 0.324%

    No Known Activations