© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    5. Llama3.1-8B-IT
    7. Resid Post - 131k
    9. 11-RESID-POST-AA
    11. 68588
    Prev
    Next
    INDEX
    Explanations

    sitting down

    np_max-act · gemini-2.0-flash

    The neuron is primarily picking up on explicit “role‐play” instruction language—tokens like “role,” “play,” “character,” etc.

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

    interactions that involve inappropriate relationships or dynamics between characters.

    oai_token-act-pair · gpt-4o-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
    wins
    -0.07
    cmd
    -0.07
    trigger
    -0.06
     Sem
    -0.06
    Ele
    -0.06
     fortn
    -0.06
     Mog
    -0.06
     TEMPLATE
    -0.06
     vicious
    -0.06
     Sed
    -0.06
    POSITIVE LOGITS
    alchemy
    0.08
     прин
    0.07
     ग
    0.07
    37
    0.06
    ')[
    0.06
     stakeholders
    0.06
     tuned
    0.06
    χές
    0.06
    ritch
    0.06
    >Delete
    0.06
    Activations Density 0.025%

    No Known Activations