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

    Neuronpedia

    APIAssistant AxisNEWCircuit TracerNEWSteerSAE EvalsExports Community BlogPrivacy & TermsContact
    © Neuronpedia 2026
    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. 126477
    Prev
    Next
    INDEX
    Explanations

    v

    np_max-act · gemini-2.0-flash

    The neuron fires on isolated subword fragments—especially single‐character or very short tokens (like “I”, “v”, “b”, “ge” or control/metadata markers)—i.e. fragmented pieces rather than whole words.

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

    tokens used for document structure, metadata, and speaker/date labels (speaker names, IDs, and numeric/date tokens).

    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
    reeting
    -0.07
    mes
    -0.06
     семей
    -0.06
     grammar
    -0.06
    463
    -0.06
     nach
    -0.06
     захворювання
    -0.06
     atIndex
    -0.06
    ifications
    -0.06
    ]!='
    -0.06
    POSITIVE LOGITS
    (Utils
    0.07
     ø
    0.06
    ायन
    0.06
     defin
    0.06
     clocks
    0.06
    [counter
    0.06
    Eval
    0.06
    ừng
    0.06
    िनक
    0.06
    ılan
    0.06
    Activations Density 0.232%

    No Known Activations