© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 6499
    Prev
    Next
    INDEX
    Explanations

    separating clauses or lists

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

    The neuron fires strongly on relatively “content-heavy” or uncommon words (the kind of nouns, verbs, or proper names that carry the main ideas of a sentence).

    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
     serta
    0.62
     можуть
    0.62
     همچنین
    0.61
     nonché
    0.61
     รวมถึง
    0.61
     sekä
    0.60
     oraz
    0.59
     등으로
    0.59
     तसेच
    0.59
     kunnen
    0.59
    POSITIVE LOGITS
     rồi
    0.62
     arms
    0.57
    alted
    0.57
     armes
    0.53
    arms
    0.53
    ocked
    0.53
    BOARD
    0.52
    flop
    0.52
     వెళ్లి
    0.52
    [
    0.52
    Activations Density 0.165%

    No Known Activations