© 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 EvalsExportsinterp-engineNEWAPI 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. 24759
    Prev
    Next
    INDEX
    Explanations

    situations where people are being asked for or struggling with requests or instructions

    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
    νÏİ
    -0.14
     елек
    -0.13
    MARY
    -0.13
    mile
    -0.13
    ANNEL
    -0.13
    >[]
    -0.12
    mary
    -0.12
    Mask
    -0.12
    maximum
    -0.12
    머
    -0.12
    POSITIVE LOGITS
     min
    1.05
     Min
    1.05
    Min
    0.94
    min
    0.93
     MIN
    0.91
    -min
    0.87
    _min
    0.84
    .min
    0.82
    (min
    0.81
     мин
    0.77
    Activations Density 0.010%

    No Known Activations