© 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. 91105
    Prev
    Next
    INDEX
    Explanations

    Dialogue

    np_max-act · gemini-2.0-flash

    This neuron activates on the special metadata/control tokens (e.g. header-ID, begin/end markers, speaker tags) that structure the conversation rather than on ordinary content words.

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

    instructions that set covert goals for roleplay—steering the conversation subtly toward a hidden agenda or eliciting help/money without stating it directly.

    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
    !?
    -0.07
    vanized
    -0.06
    Drivers
    -0.06
    .opendaylight
    -0.06
     Similar
    -0.06
     文件
    -0.06
     Ramirez
    -0.06
    apesh
    -0.05
     Sanity
    -0.05
     عبدال
    -0.05
    POSITIVE LOGITS
     Anglo
    0.07
    .pages
    0.07
    .intersection
    0.07
    ایش
    0.06
    	UN
    0.06
     panc
    0.06
    .window
    0.06
    icans
    0.06
    igma
    0.06
    .report
    0.06
    Activations Density 0.028%

    No Known Activations