© 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. Qwen3-4B
    5. 31-TRANSCODER-HP
    7. 19630
    Prev
    Next
    INDEX
    Explanations

    say "easier" or similar

    np_max-act-logits · gemini-2.0-flash
    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
    vir
    -0.25
    红线
    -0.25
    à¹Ħว
    -0.24
    oste
    -0.24
    ç¥ģ
    -0.24
    æŀģ
    -0.24
    éĤĥ
    -0.23
    dark
    -0.23
    左边
    -0.23
    é¹ŀ
    -0.23
    POSITIVE LOGITS
     easier
    0.35
    ä¸įä¸Ĭ
    0.30
     easiest
    0.29
    æĽ´å®¹æĺĵ
    0.29
    ç®ĢåįķçļĦ
    0.26
    à¹īà¸ĩ
    0.25
     lighter
    0.25
     uptime
    0.25
     less
    0.25
     grace
    0.24
    Activations Density 1.035%

    No Known Activations