© 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. 90862
    Prev
    Next
    INDEX
    Explanations

    Maintaining a state

    np_max-act · gemini-2.0-flash

    This neuron detects user instructions prompting the model (or “DAN”) to “stay in character” (i.e. direct role-play/instruction phrases to remain in character).

    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
    说
    -0.07
    µ
    -0.07
     orb
    -0.07
    Capability
    -0.07
    teen
    -0.06
    -standard
    -0.06
    посеред
    -0.06
    edom
    -0.06
    #!
    -0.06
    REAL
    -0.06
    POSITIVE LOGITS
     znam
    0.07
    иф
    0.07
     컴
    0.07
     glUniform
    0.06
    _TEMP
    0.06
     Geschichte
    0.06
     airflow
    0.06
    егод
    0.06
     khám
    0.06
    _INTER
    0.06
    Activations Density 0.003%

    No Known Activations