Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & 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. 86152
    Prev
    Next
    INDEX
    Explanations

    ’ll

    np_max-act · gemini-2.0-flash

    The neuron activates on tokens in the assistant’s standard “helping” boilerplate—words like “I’m,” “here,” “to,” “help,” “you,” “I’ll,” “do,” “my,” “best,” and “assist”—used when offering assistance.

    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
     nej
    -0.07
     Fortnite
    -0.07
     drifting
    -0.06
     SSC
    -0.06
    Equivalent
    -0.06
    NM
    -0.06
    -table
    -0.06
    .hu
    -0.06
    CPF
    -0.06
    離
    -0.06
    POSITIVE LOGITS
    (NULL
    0.06
    好像
    0.06
    sciously
    0.06
    stitutions
    0.06
     erupt
    0.06
     volunteers
    0.06
    eller
    0.06
    鄉
    0.06
     formulation
    0.06
    mun
    0.06
    Activations Density 0.019%

    No Known Activations