© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 1683
    Prev
    Next
    INDEX
    Explanations

    be careful/mindful/wary/aware

    np_acts-logits-general · gemini-2.5-flash-lite

    The neuron detects polite requests or directives—words and phrases like “please,” “note,” “make sure,” and similar polite imperatives.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_24_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     WOULD
    0.63
     wou
    0.63
     galima
    0.63
     Xml
    0.61
     prévue
    0.61
     Hunan
    0.61
     heißt
    0.61
     equals
    0.60
     accomplishes
    0.60
     lanjutan
    0.60
    POSITIVE LOGITS
     beware
    0.85
    留意
    0.81
     beachten
    0.80
    NOTE
    0.80
    注意
    0.79
    不要
    0.78
    Beware
    0.78
    อย่า
    0.78
     jangan
    0.75
    remember
    0.72
    Activations Density 0.325%

    No Known Activations