© 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. 25327
    Prev
    Next
    INDEX
    Explanations

    instances of the word "WATCH" often related to guidance or recommendations in media 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
    ãĥ¼ãĥį
    -0.17
    usercontent
    -0.15
    catch
    -0.15
    $MESS
    -0.15
    antom
    -0.14
    Ñħо
    -0.14
    irim
    -0.14
    vt
    -0.14
     Rowe
    -0.14
    æ´¥
    -0.14
    POSITIVE LOGITS
    ed
    0.27
    dog
    0.23
    ?v
    0.21
    dogs
    0.21
     Queue
    0.20
    :
    0.19
     this
    0.18
     as
    0.17
    ing
    0.17
    DOG
    0.17
    Activations Density 0.008%

    No Known Activations