© 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
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 15465
    Prev
    Next
    INDEX
    Explanations

    consideration

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

    The neuron is picking out prominent content‐labeling or header terms in legal/policy and product texts—words like “Terms,” “update,” “revision,” “new,” “Group,” “made,” or “merchandise” that typically signal section titles, formal labels, or key document headings.

    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
    я
    0.87
    ться
    0.83
    ش
    0.81
    と思いますが
    0.80
    ے
    0.73
     topping
    0.72
    ('/');
    0.72
    кову
    0.71
    ई
    0.71
    ل
    0.70
    POSITIVE LOGITS
    their
    0.93
    slides
    0.87
    not
    0.84
    that
    0.83
     yaklaşık
    0.83
     Какие
    0.82
    varchar
    0.81
     kroppen
    0.81
     kunna
    0.80
    range
    0.80
    Activations Density 0.000%

    No Known Activations