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. 6584
    Prev
    Next
    INDEX
    Explanations

    attends to the conjunctions "but" and "think" from clauses that include auxiliary verbs and are preceded by a subject in the earlier part of the sentence

    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
    0:0.15
    1:0.14
    2:0.13
    3:0.26
    4:0.11
    5:0.02
    6:0.06
    7:0.08
    Negative Logits
     Monfieur
    -0.42
     Theſe
    -0.38
     anhänger
    -0.38
    __':
    
    -0.38
     invokingState
    -0.37
    contentLoaded
    -0.36
     McKenna
    -0.36
     متعلقه
    -0.35
    IUrlHelper
    -0.34
     Metcalf
    -0.34
    POSITIVE LOGITS
    UrlResolution
    0.28
    Chham
    0.27
     Stande
    0.26
    ImageContext
    0.25
     Kenne
    0.24
     ComVisible
    0.24
    angdong
    0.24
    mrs
    0.24
     typo
    0.24
     bằng
    0.23
    Activations Density 0.184%

    No Known Activations