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. 122951
    Prev
    Next
    INDEX
    Explanations

    General conversational snippets

    np_max-act · gemini-2.0-flash

    instances where a user frames a risky or disallowed request as hypothetical/fictional or claims benign intent/ownership (e.g., “hypothetical,” “fictional,” “my own,” “researcher”) to bypass safety constraints.

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

    language that frames sensitive or harmful scenarios as hypothetical, fictional, or otherwise justified to seek permission or leniency.

    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
     알려
    -0.07
     Dept
    -0.07
    ensed
    -0.06
    ancy
    -0.06
    monthly
    -0.06
    ança
    -0.06
    (zip
    -0.06
     sample
    -0.06
     Yo
    -0.06
    ODEV
    -0.06
    POSITIVE LOGITS
    _BORDER
    0.07
    Paren
    0.07
    Mit
    0.07
    _UNSIGNED
    0.07
    _signal
    0.07
     =============================================================================↵
    0.06
     ____
    0.06
    xbc
    0.06
     Rot
    0.06
    ==='
    0.06
    Activations Density 0.241%

    No Known Activations