© 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
    7. Resid Post - 131k
    9. 11-RESID-POST-AA
    11. 41476
    Prev
    Next
    INDEX
    Explanations

    superlatives

    np_max-act · gemini-2.0-flash

    The neuron is sensitive to terms that mark ordering or extremes—especially superlatives and comparative list‐ordering words like “sweetest,” “smallest,” “from,” “to,” and similar tokens that signal ranking.

    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
     showAlert
    -0.06
     Teacher
    -0.06
    .until
    -0.06
    illance
    -0.06
    PERTIES
    -0.06
    Monday
    -0.06
    .Include
    -0.06
    ERTICAL
    -0.06
     Fri
    -0.06
     tendency
    -0.06
    POSITIVE LOGITS
     bitmask
    0.07
     occasional
    0.06
    astype
    0.06
    INLINE
    0.06
     meio
    0.06
    -office
    0.06
     VID
    0.06
     torque
    0.06
    �
    0.06
    stdbool
    0.06
    Activations Density 0.017%

    No Known Activations