© 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. 13795
    Prev
    Next
    INDEX
    Explanations

    start of phrases like "X is" or "X are"

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

    The neuron fires most strongly on tokens used in quantitative/market‐analysis contexts—things like segmentation headings (e.g. “by company,” “by country”), date ranges (e.g. “2023”), and emphasis words tied to metrics (e.g. “especially,” “peak periods”).

    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
     Saying
    0.90
     Admittedly
    0.89
     Caught
    0.84
     Archaeology
    0.82
    ႇ
    0.80
     Emergency
    0.79
     Однако
    0.79
     Parlement
    0.78
     Donations
    0.77
     Payable
    0.77
    POSITIVE LOGITS
    mul
    0.82
    lu
    0.81
    리지
    0.81
    lm
    0.80
    mical
    0.80
    bys
    0.80
    ból
    0.80
    adj
    0.78
    intestinal
    0.78
    름
    0.77
    Activations Density 0.001%

    No Known Activations