© 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
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-9B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 5482
    Prev
    Next
    INDEX
    Explanations

    code constructs related to memory management and variable types in programming

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

    low-level programming syntax, particularly related to method/function initialization and object handling in a programming language (likely Rust or a similar systems programming language).

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

    object-oriented source code, especially lines with generics, type annotations, and method/constructor signatures within strongly typed languages.

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

    programming language syntax, particularly type declarations, generic type parameters, and function signatures.

    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
    TypedDataSet
    -0.57
     youtube
    -0.56
     photoshop
    -0.51
    Autowired
    -0.50
    Youtube
    -0.49
     للمعارف
    -0.49
     NSCoder
    -0.48
    MigrationBuilder
    -0.47
     WIFI
    -0.47
     numberWith
    -0.47
    POSITIVE LOGITS
    nalités
    0.40
    vocable
    0.39
    <_>
    0.39
     necessarily
    0.36
     purport
    0.35
    étit
    0.35
    FormBorderStyle
    0.34
     případ
    0.34
    thâu
    0.33
     inevitably
    0.32
    Activations Density 0.643%

    No Known Activations