© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW 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. 26105
    Prev
    Next
    INDEX
    Explanations

    expressions of self-referential thoughts and feelings of insanity or uniqueness

    oai_token-act-pair · gpt-4o-mini
    Triggered 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
    Negative Logits
    ictim
    -0.16
    ibase
    -0.15
    ibble
    -0.14
    ارÙģ
    -0.14
    .compress
    -0.13
     misd
    -0.13
    бÑĥÑĤ
    -0.13
    ]={↵
    -0.13
    elu
    -0.13
     onCancelled
    -0.13
    POSITIVE LOGITS
     mad
    0.67
    mad
    0.61
     crazy
    0.60
     madness
    0.57
     Crazy
    0.55
     insane
    0.55
     Mad
    0.54
    Mad
    0.53
     MAD
    0.52
     craz
    0.52
    Activations Density 0.040%

    No Known Activations