© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    5. Gemma-2-9B
    7. Residual Stream - 16k
    9. 20-GEMMASCOPE-RES-16K
    11. 2815
    Prev
    Next
    INDEX
    Explanations

    HTML header tags indicating section titles or headings in the text

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    This neuron does not seem to have a clear pattern of activation. The activations are mostly zero, with a few scattered high values that do not form a coherent theme. Without more context or a clearer pattern, I cannot confidently state what this neuron is looking for.

    oai_token-act-pair · claude-3-5-haiku-20241022Triggered by @neilrathi

    HTML section headers, specifically the <h4> heading tags in markup.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408

    HTML heading tags (h4 elements).

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 20-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_20/width_16k/average_l0_68
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.20.hook_resid_post
    Hook Layer
    20
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    <h2>
    -0.60
    zzlement
    -0.55
    ()))
    -0.54
     Wikimedijinoj
    -0.53
     illustrazione
    -0.50
     "))
    -0.50
     Raim
    -0.50
    ExtendWith
    -0.48
    "))
    -0.48
     ))
    -0.47
    POSITIVE LOGITS
    <h4>
    1.32
    <h5>
    0.67
    subsubsection
    0.62
    ####
    0.60
    4
    0.54
     ####
    0.53
     Huss
    0.53
     Norbert
    0.51
    obi
    0.50
     encomp
    0.50
    Activations Density 0.041%

    No Known Activations