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. 6104
    Prev
    Next
    INDEX
    Explanations

    attends to the closing commas that follow arbitrary tokens from the opening square brackets that follow the closing square brackets

    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
    0:0.10
    1:0.15
    2:0.15
    3:0.09
    4:0.09
    5:0.11
    6:0.08
    7:0.18
    Negative Logits
     hole
    -0.27
     Graham
    -0.24
    alyptus
    -0.24
    lotten
    -0.24
     &___
    -0.24
    choss
    -0.23
     Ros
    -0.23
    ITHER
    -0.23
     desen
    -0.23
     meub
    -0.23
    POSITIVE LOGITS
    httphttps
    0.35
     Thine
    0.35
    swire
    0.34
     ddelweddau
    0.34
     referenties
    0.32
    ---*/
    0.32
     pym
    0.31
    Przypisy
    0.31
     فريبيس
    0.30
    reszcie
    0.30
    Activations Density 0.011%

    No Known Activations