© 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
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 15889
    Prev
    Next
    INDEX
    Explanations

    code, strings, or specific contexts

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

    The neuron strongly responds to structural/markup tokens (especially “<” and related boundary markers) rather than normal words.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    periods that appear at the beginning of sentences or in contexts where a new section or clause is starting.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    sentence openings at the very start of a paragraph, especially when they begin with a short prepositional phrase introducing context.

    oai_token-act-pair · gpt-5Triggered 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
    İ
    0.71
    acija
    0.70
    ays
    0.69
    ault
    0.68
    ura
    0.65
     a
    0.65
    requestId
    0.64
    ock
    0.62
    Ι
    0.62
    ;
    0.61
    POSITIVE LOGITS
     Vì
    0.89
    ൬
    0.73
    ди
    0.71
     дли
    0.71
     Например
    0.69
    CHREIB
    0.69
    ॕ
    0.69
    कारण
    0.69
    廝
    0.69
     Çünkü
    0.68
    Activations Density 0.003%

    No Known Activations