© 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. 25250
    Prev
    Next
    INDEX
    Explanations

    commas and conjunctions used in conjunction with numbers

    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
     championship
    -0.36
     championships
    -0.36
    -ch
    -0.34
     Championship
    -0.34
     Championships
    -0.33
     Chancellor
    -0.33
     chi
    -0.33
     Chicago
    -0.32
     cholesterol
    -0.32
     chamber
    -0.32
    POSITIVE LOGITS
    subcategory
    0.15
     LoÃłi
    0.15
    ÙĨÚ¯ÛĮ
    0.14
    acom
    0.14
    cup
    0.14
     Opportunities
    0.14
    æĦı
    0.14
    stvÃŃ
    0.13
    court
    0.13
    cab
    0.13
    Activations Density 0.016%

    No Known Activations