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. 6556
    Prev
    Next
    INDEX
    Explanations

    attends to causal or conditional relationships marked by "since" or "because" from following tokens indicative of explanations or justifications

    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
    1:0.09
    2:0.50
    3:0.07
    4:0.04
    5:0.02
    6:0.04
    7:0.09
    Negative Logits
    genöss
    -0.37
    はじめに
    -0.32
    ftrag
    -0.32
     langt
    -0.31
     Pre
    -0.31
    llac
    -0.30
     automo
    -0.30
     Brainstorm
    -0.30
    gnore
    -0.30
    ENCES
    -0.29
    POSITIVE LOGITS
    اریخ
    0.51
     MainAxisSize
    0.49
    IonicModule
    0.44
    AndEndTag
    0.43
    formik
    0.42
     sintético
    0.41
     surla
    0.39
    Pautan
    0.39
    љено
    0.38
     zelve
    0.38
    Activations Density 0.765%

    No Known Activations