© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    1. Home
    3. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    5. Llama3.1-8B-IT (Instruct)
    7. Resid Post - 131k
    9. 11-RESID-POST-AA
    11. 55839
    Prev
    Next
    INDEX
    Explanations

    not

    np_max-act · gemini-2.0-flash

    The neuron is keyed to informal first‐person “hedge” or disclaimer language—words like couldn’t, didn’t, made (rather than formal signals)—i.e. moments where the author is apologizing, expressing uncertainty, or qualifying their statement.

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8
    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
     Garten
    -0.06
    mongodb
    -0.06
    ht
    -0.06
    slider
    -0.06
    ών
    -0.06
    ant
    -0.06
    /web
    -0.06
    _parameters
    -0.06
    -holder
    -0.06
    ầy
    -0.06
    POSITIVE LOGITS
     valueForKey
    0.07
     arrayWith
    0.07
    向
    0.07
     untouched
    0.07
     kéo
    0.06
     никто
    0.06
     přem
    0.06
    )\↵
    0.06
     Seam
    0.06
    -ci
    0.06
    Activations Density 0.040%

    No Known Activations