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. 129361
    Prev
    Next
    INDEX
    Explanations

    code

    np_max-act · gemini-2.0-flash

    The neuron activates on hedging or advisory phrases (e.g., “Note that,” “may,” “especially,” “depending on”) used to give warnings or disclaimers.

    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
     lap
    -0.07
     dime
    -0.07
     derivative
    -0.06
    eb
    -0.06
     rabbit
    -0.06
     cool
    -0.06
     Berkeley
    -0.06
     precursor
    -0.06
     inhibit
    -0.06
     initiation
    -0.06
    POSITIVE LOGITS
    FILTER
    0.07
     suffering
    0.07
     applies
    0.06
    ære
    0.06
    .assertEqual
    0.06
     ComVisible
    0.06
     omitted
    0.06
     Cure
    0.06
    Ệ
    0.06
     Viet
    0.06
    Activations Density 0.073%

    No Known Activations