© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 9030
    Prev
    Next
    INDEX
    Explanations

    foreign language segments

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

    The neuron is essentially a “number‐finder” — it spikes on tokens that are part of numeric or quantitative expressions (dates, page‐ and volume‐numbers, measurements, prices, “24/7,” 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
    لس
    0.79
    Sessions
    0.76
    ic
    0.72
    elere
    0.72
    मेरा
    0.71
    นี่
    0.69
    वो
    0.68
    برای
    0.68
    in
    0.67
    ত্তির
    0.67
    POSITIVE LOGITS
     الاش
    1.01
     машинасы
    0.89
     определенных
    0.79
     машиналары
    0.78
     красный
    0.78
     إذ
    0.77
     изготов
    0.76
     escreveu
    0.76
     Некоторые
    0.76
     экземпля
    0.75
    Activations Density 0.002%

    No Known Activations