© 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
    2. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    3. Llama3.1-8B-IT
    4. Resid Post - 131k
    5. 27-RESID-POST-AA
    6. 27691
    Prev
    Next
    INDEX
    Explanations

    Sexual perversion

    np_max-act-logits · gemini-2.0-flash

    The neuron detects **sexual deviance**, particularly focusing on **perverted or taboo sexual behaviors**, including pedophilia, bestiality, fetishism, and extreme BDSM. It strongly activates on words like *"pervert," "disgusting," "pedo," "k9,"* and *"degrading,"* as well as explicit descriptions of morally or socially unacceptable sexual acts. This neuron appears to flag **sexually transgressive content**, especially involving **non-consensual, illegal, or extreme fetishistic themes**.

    oai_token-act-pair · deepseek-v3Triggered by @grunklewordner
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_27/trainer_1
    Dataset (Dashboard)
    Various
    Features
    131,072
    Data Type
    float32
    Hook Name
    blocks.27.hook_resid_post
    Architecture
    standard
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    _connection
    -0.07
    PP
    -0.07
    acent
    -0.07
    surface
    -0.07
     chairs
    -0.07
    AREN
    -0.07
    Chair
    -0.06
     chair
    -0.06
     chooses
    -0.06
    -commercial
    -0.06
    POSITIVE LOGITS
    ��이지
    0.06
     �
    0.06
     каче
    0.06
     إذا
    0.06
     gehört
    0.06
     sacrifices
    0.06
     Maurice
    0.06
    탕
    0.06
    삼
    0.06
     pop
    0.06
    Activations Density 0.074%

    No Known Activations