© 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. Qwen3-4B
    5. 32-TRANSCODER-HP
    7. 42380
    Prev
    Next
    INDEX
    Explanations

    to

    np_max-act-logits · gemini-2.0-flash
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    mwhanna/qwen3-4b-transcoders/layer_32.safetensors
    Prompts (Dashboard)
    16,384 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    163,840
    Data Type
    float32
    Hook Name
    blocks.32.mlp.hook_in
    Architecture
    transcoder
    Context Size
    8,192
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    åľ°
    -0.26
    sav
    -0.25
    overe
    -0.25
     Attempt
    -0.25
    çĶ³
    -0.25
    åľ°éģĵ
    -0.24
    velopment
    -0.24
    ä¹ĭäºĭ
    -0.24
     poised
    -0.24
    latex
    -0.23
    POSITIVE LOGITS
     look
    0.31
     conduct
    0.29
     one
    0.29
     catch
    0.28
     select
    0.27
     insert
    0.27
     our
    0.27
     have
    0.27
     take
    0.26
    oner
    0.26
    Activations Density 0.009%

    No Known Activations