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. 130493
    Prev
    Next
    INDEX
    Explanations

    g

    np_max-act · gemini-2.0-flash

    The neuron strongly fires on short, capitalized token fragments—i.e. 2–3-letter prefixes or abbreviations (often at the start of hyphenated or compound terms).

    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
     gần
    -0.07
    ุล
    -0.06
     [-]:
    -0.06
    -gun
    -0.06
    .GetCurrent
    -0.06
    主任
    -0.06
    horn
    -0.06
    ことも
    -0.06
    ůj
    -0.06
    ,
    -0.06
    POSITIVE LOGITS
     eigen
    0.08
     paint
    0.07
     Paint
    0.07
     posit
    0.07
     hockey
    0.06
     Push
    0.06
     offers
    0.06
     vs
    0.06
    缓
    0.06
    (norm
    0.06
    Activations Density 0.181%

    No Known Activations