© 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. 25352
    Prev
    Next
    INDEX
    Explanations

    instances of the word "defeat" and related concepts of defeatism and its implications

    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
    erer
    -0.20
    er
    -0.19
    fac
    -0.17
    ough
    -0.17
    ipt
    -0.16
      ↵  ↵
    -0.16
    erot
    -0.15
    bery
    -0.15
     lic
    -0.15
    orz
    -0.15
    POSITIVE LOGITS
    initely
    0.21
    ively
    0.20
    iveness
    0.18
    shire
    0.17
    ives
    0.17
    hower
    0.16
    asio
    0.16
    utive
    0.16
    /null
    0.16
    resher
    0.15
    Activations Density 0.032%

    No Known Activations