© 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. OpenMOSS · Llama Scope: SAEs for Llama-3.1-8B
    5. Llama3.1-8B (Base)
    7. Residual Stream
    9. 22-LLAMASCOPE-RES-32K
    11. 27131
    Prev
    Next
    INDEX
    Explanations

    imperative and action-oriented phrases related to "taking" or "removing."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With LLAMA3.1-8B @ 22-llamascope-res-32k
    Configuration
    fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L22R-8x
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    cerebras/SlimPajama-627B
    Features
    32,768
    Data Type
    bfloat16
    Hook Name
    blocks.22.hook_resid_post
    Hook Layer
    22
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    cerebras/SlimPajama-627B
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     takes
    -0.20
     taking
    -0.18
    takes
    -0.18
     take
    -0.17
    riad
    -0.16
    ime
    -0.15
     Take
    -0.15
     takeaway
    -0.15
    idth
    -0.15
    arily
    -0.15
    POSITIVE LOGITS
     advantage
    0.42
     advant
    0.25
    aways
    0.23
    uchi
    0.23
    adv
    0.23
     Advantage
    0.23
     refuge
    0.22
     care
    0.22
     advantages
    0.20
    -home
    0.20
    Activations Density 0.120%

    No Known Activations