© 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. Julius Han · Gated SAE for Llama3-8B-Instruct
    5. Llama3-8B-IT
    7. Residual Stream
    9. 25-RES-JH
    11. 43626
    Prev
    Next
    INDEX
    Explanations

    occurrences of the word "start" and its variations, indicating the theme of new beginnings or initiations

    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
                    ↵
    -0.17
    sey
    -0.17
    /to
    -0.17
    ories
    -0.16
    rung
    -0.16
     storm
    -0.15
     starting
    -0.15
     startup
    -0.15
    ulia
    -0.15
     sting
    -0.15
    POSITIVE LOGITS
    swith
    0.23
    bucks
    0.21
    utory
    0.20
    sWith
    0.20
    le
    0.19
    nings
    0.18
    -up
    0.18
    tls
    0.18
    TLS
    0.18
    ecz
    0.18
    Activations Density 0.041%

    No Known Activations