© 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. 31-TRANSCODER-HP
    7. 18863
    Prev
    Next
    INDEX
    Explanations

    period

    np_max-act-logits · gemini-2.0-flash

    code snippets and technical file extensions (like `.txt`, `.php`, `.json`, etc.) in programming-related queries.

    oai_token-act-pair · deepseek-v3Triggered by @finnwit

    Nouns, punctuation marks, and common connectors following user queries or code snippets, often preceding or following technical terms or instructions.

    eleuther_acts_top20 · deepseek-v3Triggered by @finnwit
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    mwhanna/qwen3-4b-transcoders/layer_31.safetensors
    Prompts (Dashboard)
    16,384 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    163,840
    Data Type
    float32
    Hook Name
    blocks.31.mlp.hook_in
    Architecture
    transcoder
    Context Size
    8,192
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Large
    -0.42
     Simple
    -0.39
     Unique
    -0.39
     Better
    -0.39
     Highly
    -0.38
     Helpful
    -0.37
     Lesser
    -0.36
     Huge
    -0.36
     ÐijолÑĮÑĪ
    -0.36
    Minimal
    -0.35
    POSITIVE LOGITS
     The
    0.35
    The
    0.35
    	The
    0.32
    >The
    0.31
    /The
    0.30
    nThe
    0.30
    _THE
    0.26
    _REGS
    0.26
    æĭ¿çĿĢ
    0.25
    .The
    0.25
    Activations Density 35.866%

    No Known Activations