© 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. 22-GEMMASCOPE-TRANSCODER-16K
    7. 5628
    Prev
    Next
    INDEX
    Explanations

    a combination of several patterns including negative contexts like "undesired", programming terms, personal pronouns followed by contracted forms of auxiliary verbs and words relating to social media

    oai_token-act-pair · gemini-2.0-flash

    code and technical language

    np_max-act-logits · gemini-2.0-flash
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2b-pt-transcoders/layer_22/width_16k/average_l0_15
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.22.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
     want
    -2.38
    want
    -2.25
     wants
    -2.23
     Want
    -2.19
     wanted
    -2.17
     wanting
    -2.06
    Want
    -1.95
    wants
    -1.94
     Wanted
    -1.94
    wanted
    -1.94
    POSITIVE LOGITS
     flywheel
    0.40
    thoven
    0.39
    medades
    0.38
    isContained
    0.37
     ANTONIO
    0.37
    nean
    0.36
    lamy
    0.36
     tel
    0.35
    subpackage
    0.35
    局
    0.35
    Activations Density 1.154%

    No Known Activations