© 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. 55170
    Prev
    Next
    INDEX
    Explanations

    references to witches and related fantastical elements

    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
     Civilization
    -0.15
    ynos
    -0.15
    orges
    -0.14
    iami
    -0.14
    rong
    -0.14
     jungle
    -0.14
    inki
    -0.14
    xygen
    -0.14
     Flags
    -0.13
    ierge
    -0.13
    POSITIVE LOGITS
     witch
    0.49
     witches
    0.46
     Witch
    0.44
    witch
    0.42
     Salem
    0.32
     sor
    0.29
     Spells
    0.27
     spells
    0.27
    itches
    0.27
     spell
    0.26
    Activations Density 0.021%

    No Known Activations