© 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. 31-GEMMASCOPE-RES-16K
    6. 3843
    Prev
    Next
    INDEX
    Explanations

    formal conclusions and decision statements

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

    formatting elements like whitespace, line breaks, and structural markers in documents.

    oai_token-act-pair · claude-4-5-haikuTriggered by @emiglarou

    line breaks or whitespace between sections in formal legal, technical, or structured documents.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    special whitespace/Unicode formatting artifacts and structural separators that indicate highly formatted or copy-pasted document sections, often from legal or tabular content.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 31-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_31/width_16k/average_l0_114
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.31.hook_resid_post
    Hook Layer
    31
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    +:+
    -0.41
    <strong>
    -0.34
    ↵↵↵
    -0.33
    rawdę
    -0.32
    addOn
    -0.32
    pherd
    -0.30
    iterranée
    -0.29
     kaynağından
    -0.29
    lgari
    -0.29
    不觉
    -0.29
    POSITIVE LOGITS
    出版年
    0.68
     дописавши
    0.65
     Vikipedi
    0.64
     Италијани
    0.62
     برانيه
    0.61
    AnimationsModule
    0.60
    expandindo
    0.58
     HasFactory
    0.57
     disambiguazione
    0.57
    WithIOException
    0.56
    Activations Density 0.246%

    No Known Activations