© 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. 40749
    Prev
    Next
    INDEX
    Explanations

    negative or cautionary phrases and expressions related to safety or directives

    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
    ennon
    -0.14
    ácil
    -0.14
    slideUp
    -0.14
     downfall
    -0.14
    ADIUS
    -0.14
    ÏĮÏĤ
    -0.14
    isis
    -0.13
    otime
    -0.13
    adr
    -0.13
    cntl
    -0.13
    POSITIVE LOGITS
     touch
    0.36
     Touch
    0.32
    touch
    0.30
    Touch
    0.29
     TOUCH
    0.28
     touched
    0.28
     touches
    0.27
    -touch
    0.27
     touching
    0.27
    _touch
    0.22
    Activations Density 0.043%

    No Known Activations