© 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. 28462
    Prev
    Next
    INDEX
    Explanations

    the word "why" in various contexts

    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
    .DropDown
    -0.17
    âĹİ
    -0.15
    Ïĥκε
    -0.14
    âĸį
    -0.14
     WHETHER
    -0.14
    æ¹¾
    -0.14
    .Bounds
    -0.14
    ...↵↵↵↵
    -0.14
    ogui
    -0.14
    adu
    -0.13
    POSITIVE LOGITS
    /how
    0.39
     certain
    0.28
     they
    0.28
    soever
    0.27
     we
    0.27
     exactly
    0.27
     bother
    0.24
     it
    0.23
     this
    0.23
     he
    0.20
    Activations Density 0.049%

    No Known Activations