© 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. 18-TRANSCODER-HP
    7. 89215
    Prev
    Next
    INDEX
    Explanations

    code and file paths

    np_max-act · gemini-2.0-flash

    technical/coding

    np_max-act-logits · deepseek-v3Triggered by @finnwit

    a

    np_max-act · deepseek-v3Triggered by @finnwit

    Tokens that indicate code syntax errors, missing dependencies, or incorrect implementations, often appearing near delimiters in programming contexts.

    eleuther_acts_top20 · deepseek-v3Triggered by @finnwit
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    mwhanna/qwen3-4b-transcoders/layer_18.safetensors
    Prompts (Dashboard)
    16,384 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    163,840
    Data Type
    float32
    Hook Name
    blocks.18.mlp.hook_in
    Architecture
    transcoder
    Context Size
    8,192
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    ire
    -0.29
    ann
    -0.27
     TOUCH
    -0.26
    ä¸ĢåIJĮ
    -0.26
    iring
    -0.25
    ieur
    -0.25
    rive
    -0.25
    ari
    -0.24
    NUM
    -0.24
    NM
    -0.24
    POSITIVE LOGITS
     unrelated
    0.32
    æĹłåħ³
    0.26
    ä¸įæĪIJ
    0.26
    ç°§
    0.26
    ä¸ŃåİŁ
    0.25
    CurrentValue
    0.25
     bank
    0.25
    âĬĦ
    0.25
    éĵ¶è¡Į
    0.25
    >Delete
    0.25
    Activations Density 27.674%

    No Known Activations