© 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. OpenMOSS · Llama Scope: SAEs for Llama-3.1-8B
    5. Llama3.1-8B (Base)
    7. Residual Stream
    9. 21-LLAMASCOPE-RES-32K
    11. 7027
    Prev
    Next
    INDEX
    Explanations

    phrases related to protection and safeguarding against various threats

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With LLAMA3.1-8B @ 21-llamascope-res-32k
    Configuration
    fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L21R-8x
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    cerebras/SlimPajama-627B
    Features
    32,768
    Data Type
    bfloat16
    Hook Name
    blocks.21.hook_resid_post
    Hook Layer
    21
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    cerebras/SlimPajama-627B
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     somewhere
    -0.15
    ewhere
    -0.14
     winners
    -0.14
    ucha
    -0.14
    ehler
    -0.14
    ailer
    -0.14
    .issue
    -0.13
    èĥ¶
    -0.13
    ahl
    -0.13
     máu
    -0.13
    POSITIVE LOGITS
     being
    0.27
     harm
    0.25
     attack
    0.24
     scrutiny
    0.23
     becoming
    0.23
     harms
    0.22
     attacks
    0.22
    æĿ¥èĩª
    0.21
     Being
    0.20
     influences
    0.20
    Activations Density 0.125%

    No Known Activations