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-2-9B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 0
    Prev
    Next
    INDEX
    Explanations

    technical terminology related to programming structures and operations

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

    code-related computational operations, particularly involving indexing, loops, and mathematical calculations in programming languages like C++ and Python.

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

    The text features code snippets interspersed with mathematical notation and discussions about programming and formulations in different languages, indicating a technical context focused on programming, math problems, and algorithm implementation. Common patterns include embedded code formatting and comparisons in algorithms, alongside commentary on potential issues and concepts.

    eleuther_acts_top20 · gpt-4o-miniTriggered by @neilrathi

    mathematical and programming expressions, especially those involving calculations or iterations.

    oai_token-act-pair · gpt-4oTriggered by @neilrathi
    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
     betweenstory
    -0.64
     незавершена
    -0.59
    WEBPACK
    -0.56
    <unused55>
    -0.55
     fashiola
    -0.54
     model
    -0.54
    <unused79>
    -0.54
    <unused23>
    -0.54
    <unused14>
    -0.54
    [@BOS@]
    -0.54
    POSITIVE LOGITS
    OGND
    0.46
    kháu
    0.45
     Juifs
    0.33
    ConstraintMaker
    0.33
     lendemain
    0.30
     AssemblyCulture
    0.29
    цездатний
    0.29
    UNICIP
    0.28
    énario
    0.28
    MessageTagHelper
    0.27
    Activations Density 0.258%

    No Known Activations