© 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. 63689
    Prev
    Next
    INDEX
    Explanations

    phrases and terms indicating safety and security

    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
     safety
    -0.20
    soever
    -0.20
     Safety
    -0.18
    ls
    -0.17
    ship
    -0.17
    loth
    -0.17
    Safety
    -0.16
    inous
    -0.16
    afety
    -0.15
    son
    -0.15
    POSITIVE LOGITS
    -guard
    0.23
    AreaView
    0.21
     enough
    0.20
    keeping
    0.19
     harbor
    0.18
     Harbor
    0.18
    (r
    0.16
    ably
    0.16
     haven
    0.16
     hav
    0.16
    Activations Density 0.022%

    No Known Activations