© 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. 29742
    Prev
    Next
    INDEX
    Explanations

    weapons

    np_max-act · gemini-2.0-flash

    requests or mentions of illegal or harmful activities, particularly involving instructions or discussions on creating or using dangerous substances or methods.

    oai_token-act-pair · gpt-4oTriggered by @tcai

    The main thing this neuron does is detect user requests for step-by-step instructions on how to carry out illicit or harmful activities.

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

    phrases related to dangerous or harmful activities.

    oai_token-act-pair · gpt-4o-miniTriggered by @xinyanhu8

    requests for explicit instructions to commit harm, illegal activity, or create malicious/unsafe tools (e.g., viruses, explosives, poisoning, tampering).

    oai_token-act-pair · gpt-5-miniTriggered 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
     Employ
    -0.06
     센
    -0.06
    ований
    -0.06
    vestment
    -0.06
    dismiss
    -0.06
    �
    -0.06
    hf
    -0.06
     threaded
    -0.06
    Desk
    -0.06
    統
    -0.06
    POSITIVE LOGITS
    978
    0.07
    悠
    0.07
     AVR
    0.07
    ------------------------------------------------------------------------------------------------
    0.07
     durability
    0.07
     selectable
    0.07
     Bib
    0.06
    ephir
    0.06
     lake
    0.06
    (dim
    0.06
    Activations Density 1.229%

    No Known Activations