© 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. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    5. Llama3.1-8B-IT (Instruct)
    7. Resid Post - 131k
    9. 11-RESID-POST-AA
    11. 13496
    Prev
    Next
    INDEX
    Explanations

    Code/technical documents

    np_max-act · gemini-2.0-flash

    The neuron strongly activates on tokens that appear in the system-level or header metadata (role labels, directive keywords, and other structural prompt instructions), i.e. it flags bits of the system prompt rather than user or assistant content.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    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
    rames
    -0.07
    .btn
    -0.07
    .Brand
    -0.06
     Grande
    -0.06
     곧
    -0.06
    .ONE
    -0.06
     SV
    -0.06
    .fm
    -0.06
    ahren
    -0.06
    point
    -0.06
    POSITIVE LOGITS
    divid
    0.06
     Pepsi
    0.06
    WP
    0.06
     checker
    0.06
     pada
    0.06
    iano
    0.06
    ("/{
    0.06
     compelling
    0.06
    .UserService
    0.06
     цього
    0.06
    Activations Density 0.013%

    No Known Activations