© 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. 7548
    Prev
    Next
    INDEX
    Explanations

    specifications and descriptions

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

    The neuron is selectively activating on tokens that come from structured metadata or reference‐style material—things like table entries, numeric statistics, citation fragments (author initials, years), URLs/domains, and other non-narrative tokens.

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

    the word "are" in formal or technical contexts.

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

    short all-caps abbreviations and initialisms (including country/domain code fragments), especially those appearing in citations and URLs.

    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
    mmmm
    0.79
     Chernobyl
    0.78
     litmus
    0.78
     并
    0.76
     redeeming
    0.74
     千
    0.72
     возраст
    0.72
    %、
    0.71
    ıyoruz
    0.71
     blowout
    0.71
    POSITIVE LOGITS
     skup
    0.75
    )\$(
    0.72
     میان
    0.72
    TER
    0.71
     Daly
    0.70
    ard
    0.69
    ̣ng
    0.68
     Ral
    0.68
    नि
    0.67
     فيما
    0.66
    Activations Density 0.001%

    No Known Activations