© 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
    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. 90719
    Prev
    Next
    INDEX
    Explanations

    Internet chat punctuation

    np_max-act · gemini-2.0-flash

    The neuron is triggered by the model’s internal header‐delimiter tokens that mark the beginning of a new speaker or system block (e.g. `<|start_header_id|>`).

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

    tokens that begin assistant responses (especially greeting/intro phrases like "Hello! How can I help you today?").

    oai_token-act-pair · gpt-5-miniTriggered 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
     resource
    -0.08
    Omega
    -0.07
    imbabwe
    -0.07
    γχ
    -0.07
     ideal
    -0.07
    ivery
    -0.06
     Ch
    -0.06
     Filter
    -0.06
    (compare
    -0.06
    ipa
    -0.06
    POSITIVE LOGITS
    父亲
    0.07
     попыт
    0.07
     jTextField
    0.07
     περί
    0.06
    htm
    0.06
     pupper
    0.06
    ництво
    0.06
    =%.
    0.06
     JMenuItem
    0.06
     responder
    0.06
    Activations Density 0.028%

    No Known Activations