© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Julius Han · Gated SAE for Llama3-8B-Instruct
    5. Llama3-8B-IT
    7. Residual Stream
    9. 25-RES-JH
    11. 55256
    Prev
    Next
    INDEX
    Explanations

    the word "do" in various contexts, indicating an emphasis on actions or tasks

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    Juliushanhanhan/llama-3-8b-it-res/blocks.25.hook_resid_post
    Features
    65,536
    Data Type
    float32
    Hook Name
    blocks.25.hook_resid_post
    Hook Layer
    25
    Architecture
    gated
    Context Size
    1,024
    Dataset
    Juliushanhanhan/openwebtext-1b-llama3-tokenized-cxt-1024
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    doing
    -0.23
    pher
    -0.20
     doing
    -0.19
    noon
    -0.19
    did
    -0.17
     Doing
    -0.16
    unga
    -0.16
    do
    -0.16
    Doing
    -0.16
    ting
    -0.16
    POSITIVE LOGITS
     not
    0.25
    ctype
    0.24
    ctest
    0.23
    ctr
    0.23
    oms
    0.22
    cket
    0.22
    xor
    0.22
     NOT
    0.22
    zens
    0.22
    ze
    0.22
    Activations Density 0.149%

    No Known Activations