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. 128943
    Prev
    Next
    INDEX
    Explanations

    excerpts from texts

    np_max-act · gemini-2.0-flash

    This neuron detects the special system token sequence marking the transition from the header into the assistant’s reply (i.e. the end-of-header or start-of-assistant‐output marker).

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

    the beginning of an assistant’s reply, especially the initial tokens that mark or kick off a story-like response.

    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
     pj
    -0.07
    當
    -0.07
    Person
    -0.07
     shades
    -0.07
     Utility
    -0.06
    erral
    -0.06
     pdu
    -0.06
     Ig
    -0.06
     negativity
    -0.06
     провод
    -0.06
    POSITIVE LOGITS
     ====
    0.07
    dehy
    0.07
     egregious
    0.06
    SSH
    0.06
     国家
    0.06
    raph
    0.06
     tempting
    0.06
     rodiny
    0.06
     latest
    0.06
    Service
    0.06
    Activations Density 0.031%

    No Known Activations