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. 6235
    Prev
    Next
    INDEX
    Explanations

    attends to phrases or tokens that express high intensity or significance from phrases that follow without any specific context or directives

    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
    0:0.06
    1:0.18
    2:0.07
    3:0.12
    4:0.15
    5:0.27
    6:0.07
    7:0.04
    Negative Logits
    évaluateur
    -0.31
    ✨:
    -0.28
    Portail
    -0.27
     $_"
    -0.27
     Conj
    -0.27
    الإنجليزية
    -0.26
     RIPRODUZIONE
    -0.26
    Последние
    -0.26
     समीक्षक
    -0.26
    onedDateTime
    -0.25
    POSITIVE LOGITS
     Meksiku
    0.38
    PreferredItem
    0.37
    ContentAsync
    0.32
    BufferException
    0.30
    HostException
    0.30
    MLLoader
    0.29
    tawesome
    0.29
    DataAnnotations
    0.28
    ConstraintMaker
    0.27
    ároz
    0.27
    Activations Density 0.013%

    No Known Activations