© 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 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. 14298
    Prev
    Next
    INDEX
    Explanations

    phrases emphasizing the act of stopping or preventing actions, particularly in contexts related to conflict, danger, and truth

    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
    erd
    -0.18
    عاÙĨ
    -0.15
    ãģ»ãģĨ
    -0.14
    ctl
    -0.14
     explos
    -0.14
    datas
    -0.14
    418
    -0.14
     æ¡
    -0.14
    еÑĢг
    -0.14
    contri
    -0.14
    POSITIVE LOGITS
     progress
    0.32
     momentum
    0.28
     flow
    0.28
     further
    0.27
     progression
    0.25
     tide
    0.24
     advance
    0.24
     spread
    0.23
     presses
    0.23
    -progress
    0.23
    Activations Density 0.034%

    No Known Activations