© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    3. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    5. Gemma-2-9B
    7. Residual Stream - 16k
    9. 36-GEMMASCOPE-RES-16K
    11. 5114
    Prev
    Next
    INDEX
    Explanations

    references to discussions of secrets and information retention

    oai_token-act-pair · gpt-4o-mini
    Triggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 36-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_36/width_16k/average_l0_120
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.36.hook_resid_post
    Hook Layer
    36
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Negative Logits
     said
    -1.02
     says
    -0.99
    says
    -0.93
     Says
    -0.86
    Said
    -0.83
     obviously
    -0.82
    said
    -0.82
     reportedly
    -0.81
    Says
    -0.81
     Obviously
    -0.75
    POSITIVE LOGITS
     pihaknya
    0.71
     könne
    0.51
    才是
    0.50
     efectivamente
    0.49
    之所以
    0.48
     وأن
    0.47
    ValueStyle
    0.47
     puissent
    0.47
     indeed
    0.46
     désormais
    0.46
    Activations Density 1.302%

    No Known Activations