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

    Neuronpedia

    APIAssistant AxisNEWCircuit 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. 6492
    Prev
    Next
    INDEX
    Explanations

    attends to specific legal references and citations from surrounding tokens

    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
    0:0.08
    1:0.10
    2:0.17
    3:0.14
    4:0.13
    5:0.17
    6:0.12
    7:0.06
    Negative Logits
     myſelf
    -0.44
    ſelf
    -0.43
     themſelves
    -0.41
     himſelf
    -0.39
     preſent
    -0.35
     Efq
    -0.34
     houſe
    -0.34
    ſelves
    -0.34
    Искәрмәләр
    -0.33
     beſt
    -0.33
    POSITIVE LOGITS
    MessageTagHelper
    0.28
    0.26
     ]
    
    0.25
     t
    0.24
     '
    0.24
     препратки
    0.23
    DockStyle
    0.23
     ]
    0.23
    ]!=
    0.23
    </blockquote>
    0.22
    Activations Density 0.036%

    No Known Activations