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. Attention Out - 16k
    5. 10-GEMMASCOPE-ATT-16K
    6. 6094
    Prev
    Next
    INDEX
    Explanations

    attends to empty or placeholder tokens from tokens that are also empty or placeholder entities

    oai_attention-head · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 10-gemmascope-att-16k
    Configuration
    google/gemma-scope-2b-pt-att/layer_10/width_16k/average_l0_70
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.10.attn.hook_z
    Hook Layer
    10
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Head Attr Weights
    0:0.12
    1:0.17
    2:0.12
    3:0.09
    4:0.12
    5:0.14
    6:0.11
    7:0.09
    Negative Logits
    تقاوى
    -0.32
     referenties
    -0.31
    EDEFAULT
    -0.30
    saraba
    -0.30
    InjectAttribute
    -0.29
    EndContext
    -0.28
     ModelExpression
    -0.27
    Referências
    -0.27
     AppComponent
    -0.26
    bigliamento
    -0.26
    POSITIVE LOGITS
    μβρίου
    0.30
    /**
    0.29
    scrapy
    0.29
    hopped
    0.26
    unnitel
    0.25
    Tru
    0.24
    ContentAsync
    0.24
    gouv
    0.23
    shank
    0.23
    ttemberg
    0.23
    Activations Density 0.008%

    No Known Activations