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

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 12970
    Prev
    Next
    INDEX
    Explanations

    still / despite

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

    The neuron fires on content‐bearing words—especially relatively low‐frequency nouns and adjectives that carry real meaning (e.g. “price,” “cashmere,” “arctic,” “fine,” “yarn,” etc.).

    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
     قلي
    0.70
    清理
    0.69
     trochu
    0.68
     biraz
    0.68
    ätzen
    0.66
    좀
    0.66
    几个
    0.66
     vários
    0.66
    শাসনের
    0.65
     alguns
    0.64
    POSITIVE LOGITS
     tetap
    1.36
     despite
    1.24
    꿋
    1.19
     regardless
    1.17
    regardless
    1.15
    despite
    1.14
     несмотря
    1.13
     vẫn
    1.11
    即便
    1.08
    still
    1.06
    Activations Density 1.399%

    No Known Activations