© 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. 41935
    Prev
    Next
    INDEX
    Explanations

    for

    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
    uses
    -0.29
    éĩį
    -0.28
     Pars
    -0.27
    peg
    -0.26
     realize
    -0.25
    ockey
    -0.25
     responsible
    -0.25
     realise
    -0.24
    аÑĢам
    -0.24
    è´Łè´£
    -0.24
    POSITIVE LOGITS
     allowing
    0.56
    allow
    0.38
    .allow
    0.36
    åħģ许
    0.36
     permitting
    0.33
     abusing
    0.33
     introducing
    0.31
    Allow
    0.30
     recommending
    0.30
     letting
    0.30
    Activations Density 0.003%

    No Known Activations