© 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. 14796
    Prev
    Next
    INDEX
    Explanations

    negative or critical phrases related to consequences or demands

    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
    orphism
    -0.15
    aira
    -0.15
    -looking
    -0.14
    éra
    -0.14
    еко
    -0.14
    reed
    -0.14
    év
    -0.13
    --
    -0.13
    arias
    -0.13
    swire
    -0.13
    POSITIVE LOGITS
    izon
    0.14
    readcr
    0.14
    Ø©
    0.13
    ————————
    0.13
    ãģĹãģªãģĦ
    0.13
    IIIK
    0.13
    оÑĢаÑı
    0.13
     cum
    0.12
    HeaderCode
    0.12
    isma
    0.12
    Activations Density 0.003%

    No Known Activations