© 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. 30574
    Prev
    Next
    INDEX
    Explanations

    references to violent actions and specific attack methods involving sharp objects

    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
     cannon
    -0.17
    çĤ®
    -0.16
     Bracket
    -0.16
     shoot
    -0.15
     Fired
    -0.15
    aur
    -0.14
    iram
    -0.14
    reu
    -0.14
     cannons
    -0.14
     Sherman
    -0.14
    POSITIVE LOGITS
     knife
    0.53
     knives
    0.51
     Knife
    0.47
     blade
    0.47
     blades
    0.43
    knife
    0.41
    Knife
    0.40
    blade
    0.37
    åĪĢ
    0.36
     Blades
    0.36
    Activations Density 0.032%

    No Known Activations