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-2-9B
    4. Attention Out - 16k
    5. 0-GEMMASCOPE-ATT-16K
    6. 1242
    Prev
    Next
    INDEX
    Explanations

    attends to the word "que" from tokens that are framed with the question format

    oai_attention-head · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-9b-pt-att/layer_0/width_16k/average_l0_61
    Prompts (Dashboard)
    16,384 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.0.attn.hook_z
    Hook Layer
    0
    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
    1:0.04
    2:0.02
    3:0.02
    4:0.48
    5:0.06
    6:0.05
    7:0.05
    8:0.01
    9:0.02
    10:0.02
    11:0.02
    12:0.03
    13:0.03
    14:0.02
    15:0.01
    Negative Logits
     que
    -0.74
     en
    -0.57
     y
    -0.56
     et
    -0.55
     para
    -0.54
     la
    -0.52
     con
    -0.51
     por
    -0.51
     de
    -0.50
     =
    -0.50
    POSITIVE LOGITS
     myſelf
    0.87
     itſelf
    0.85
     Anſ
    0.82
     feroit
    0.82
     auroit
    0.81
     zijne
    0.80
     themſelves
    0.79
     Reſ
    0.77
     avoient
    0.77
     Houſe
    0.76
    Activations Density 0.028%

    No Known Activations