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. 125618
    Prev
    Next
    INDEX
    Explanations

    ar

    np_max-act · gemini-2.0-flash

    The neuron fires strongly on mentions of “vampire” or vampire‐related context.

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

    This neuron responds to sexual or sexualized content—especially first-person/confessional and taboo roleplay themes (arousal, attraction, incestuous or drugging scenarios).

    oai_token-act-pair · gpt-5-miniTriggered by @vetterc0

    descriptions of sexual or highly romantic scenarios—often taboo or fetishized roleplay—emphasizing attraction, arousal, and intimate feelings.

    oai_token-act-pair · gpt-5Triggered 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
     мяс
    -0.06
    ximity
    -0.06
     spre
    -0.06
    =message
    -0.06
     марш
    -0.06
     Liberty
    -0.06
     trash
    -0.06
    ulous
    -0.06
     Admir
    -0.06
    ,status
    -0.06
    POSITIVE LOGITS
    Eq
    0.07
    oron
    0.06
    Finite
    0.06
     classifications
    0.06
     bufsize
    0.06
     heroes
    0.06
     �
    0.06
     forbid
    0.06
    General
    0.06
    檢
    0.06
    Activations Density 0.240%

    No Known Activations