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-2-2B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 10695
    Prev
    Next
    INDEX
    Explanations

    punctuation marks and variation in sentence endings

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

    Follows periods at the end of sentences

    np_acts-logits-general · gemini-2.0-flash

    German, Danish, Russian, Slavic, English words followed by specific punctuation or function words

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 20-gemmascope-res-16k
    Configuration
    google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71
    Prompts (Dashboard)
    36,864 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
    aka
    -0.83
     incentiv
    -0.80
     showcasing
    -0.77
     aka
    -0.77
     leveraging
    -0.71
    AKA
    -0.71
     AKA
    -0.69
    Additionally
    -0.68
     للاسماء
    -0.68
    FYI
    -0.66
    POSITIVE LOGITS
     Daß
    0.81
     doubtless
    0.69
     forthwith
    0.60
    UserScript
    0.59
     scarcely
    0.59
     ибо
    0.59
     läßt
    0.58
     Надо
    0.57
     daß
    0.54
     muß
    0.53
    Activations Density 0.593%

    No Known Activations