Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    3. Llama3.1-8B-IT (Instruct)
    4. Resid Post - 131k
    5. 11-RESID-POST-AA
    6. 26778
    Prev
    Next
    INDEX
    Explanations

    Labeled data

    np_max-act · gemini-2.0-flash

    The neuron fires primarily on the special end‐of‐header marker token (i.e. `<|end_header_id|>`), indicating it’s detecting the boundary between header and body text.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8

    indicators of the assistant’s response sections, especially header/meta tokens and structured formatting like tables or lists.

    oai_token-act-pair · gpt-5Triggered by @vetterc0
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_11/trainer_1
    Dataset (Dashboard)
    Various
    Features
    131,072
    Data Type
    float32
    Hook Name
    blocks.11.hook_resid_post
    Architecture
    standard
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     Richmond
    -0.06
    ثمان
    -0.06
    하우
    -0.06
    Components
    -0.06
    loub
    -0.06
     هواپیم
    -0.06
     กรก
    -0.06
     ukon
    -0.06
    uclear
    -0.06
    (@"
    -0.06
    POSITIVE LOGITS
     وان
    0.07
    ##_
    0.06
     trailer
    0.06
     даль
    0.06
     contrary
    0.06
     القدم
    0.06
     �
    0.06
     تجربه
    0.06
     oats
    0.06
     意
    0.06
    Activations Density 0.053%

    No Known Activations