Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & 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. 6588
    Prev
    Next
    INDEX
    Explanations

    attends to "channel" and "RS" tokens from "channel" and "RS" tokens respectively at the start or adjacent positions in sentences

    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
    0:0.04
    1:0.10
    2:0.05
    3:0.03
    4:0.20
    5:0.46
    6:0.03
    7:0.04
    Negative Logits
    })));
    -0.24
    WindowConstants
    -0.22
     lucifer
    -0.22
     geschlossen
    -0.22
     야
    -0.21
    OrEqual
    -0.21
    among
    -0.20
    ără
    -0.20
     Bangor
    -0.20
    uled
    -0.20
    POSITIVE LOGITS
    DoubleQuotes
    0.29
    المكان
    0.29
    Билгалдахарш
    0.28
    Зноскі
    0.28
    COUVER
    0.28
     estekak
    0.28
     ProtoMessage
    0.27
     utafitiHapana
    0.27
    AnchorTagHelper
    0.27
    Aholisi
    0.27
    Activations Density 0.566%

    No Known Activations