© 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. 42046
    Prev
    Next
    INDEX
    Explanations

    the use of transitional phrases like "but" and "and" that indicate shifts in argument or contrast

    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
    iales
    -0.17
    ĺìĿ´
    -0.17
     therefore
    -0.16
     but
    -0.16
     Therefore
    -0.16
     maar
    -0.15
     moreover
    -0.15
    isphere
    -0.15
    aby
    -0.14
    hell
    -0.14
    POSITIVE LOGITS
     although
    0.26
     though
    0.22
     this
    0.20
     aunque
    0.20
    this
    0.19
    although
    0.19
     these
    0.18
     oh
    0.17
     such
    0.17
     alongside
    0.16
    Activations Density 0.029%

    No Known Activations