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-2-2B
    4. Residual Stream - 16k
    5. 13-GEMMASCOPE-RES-16K
    6. 14334
    Prev
    Next
    INDEX
    Explanations

    instances of the word "hold" and its variations

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 13-gemmascope-res-16k
    Configuration
    google/gemma-scope-2b-pt-res/layer_13/width_16k/average_l0_84
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.13.hook_resid_post
    Hook Layer
    13
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    genossen
    -0.79
    ucar
    -0.72
    EEN
    -0.72
     bác
    -0.70
    PyExc
    -0.69
    ccn
    -0.68
    rrggbb
    -0.68
     routes
    -0.67
     Pathways
    -0.67
     jSON
    -0.67
    POSITIVE LOGITS
    holds
    1.41
     Hold
    1.39
     Holds
    1.38
     holds
    1.37
     hold
    1.37
     Holden
    1.35
     HOLD
    1.32
    Hold
    1.28
    HOLD
    1.27
     holding
    1.26
    Activations Density 0.098%

    No Known Activations