© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports 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. 33151
    Prev
    Next
    INDEX
    Explanations

    phrases that indicate the presence of unauthorized or deceptive actions

    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
    •Wall Street Topics
    fraud

    No Comments

    Negative Logits
    chwitz
    -0.15
     już
    -0.14
    ola
    -0.14
    Ä©nh
    -0.14
    URRE
    -0.13
     threats
    -0.13
     survival
    -0.13
     violations
    -0.13
    iration
    -0.13
    ÄĻż
    -0.13
    POSITIVE LOGITS
     technical
    0.23
     somehow
    0.22
     software
    0.21
    technical
    0.20
    Technical
    0.20
     fault
    0.19
     Technical
    0.19
     glitch
    0.19
    fault
    0.17
    Fault
    0.16
    Activations Density 0.054%

    No Known Activations