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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-2B
    5. 19-GEMMASCOPE-TRANSCODER-16K
    7. 8525
    Prev
    Next
    INDEX
    Explanations

    the word "maintain" and related words and topics combined with evaluative language, such as "consistent", "safety", or "threats".

    oai_token-act-pair · gemini-2.0-flash

    Keeping something at a level

    np_max-act-logits · gemini-2.0-flash
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2b-pt-transcoders/layer_19/width_16k/average_l0_12
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.19.ln2.hook_normalized
    Architecture
    jumprelu_transcoder
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    Vidite
    -0.56
     Rumuni
    -0.52
    utuhkan
    -0.48
    mallows
    -0.47
     FormBuilder
    -0.44
    iosamente
    -0.42
     fédéral
    -0.42
     lenker
    -0.42
     formando
    -0.42
     pewno
    -0.42
    POSITIVE LOGITS
    aarrggbb
    0.75
    GEBURTSDATUM
    0.61
    <bos>
    0.61
    GeneratedMessage
    0.60
    ernes
    0.60
    новништво
    0.59
    Tembelea
    0.59
    WriteTagHelper
    0.57
     iconLine
    0.57
    rxjs
    0.57
    Activations Density 5.301%

    No Known Activations