© 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. 91569
    Prev
    Next
    INDEX
    Explanations

    Code/technical documentation

    np_max-act · gemini-2.0-flash

    The neuron activates on Spanish instructions for correcting grammar (e.g. words like corregir, errores gramaticales, oración), so it’s looking for Spanish-language grammar‐correction directives.

    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
     discount
    -0.07
    OA
    -0.07
     Technician
    -0.06
     jacket
    -0.06
    _INCLUDED
    -0.06
    MAN
    -0.06
     Ihr
    -0.06
    enaries
    -0.06
     vaccinated
    -0.06
    วม
    -0.06
    POSITIVE LOGITS
    (Request
    0.07
    >Note
    0.07
     UNIQUE
    0.06
    .TODO
    0.06
     ceil
    0.06
    lege
    0.06
    _TAC
    0.06
     poměr
    0.06
    .Enums
    0.06
    靈
    0.06
    Activations Density 0.066%

    No Known Activations