© 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. 31-GEMMASCOPE-2-RES-16K
    7. 6425
    Prev
    Next
    INDEX
    Explanations

    model

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

    The neuron activates strongly on punctuation marks—especially commas and periods—marking sentence or clause boundaries.

    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_31_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
     prü
    0.80
     unorthodox
    0.80
     reptiles
    0.75
     irresponsible
    0.72
     swallowed
    0.71
    <s>
    0.70
     beetles
    0.70
     frowning
    0.69
     manusia
    0.68
     cláss
    0.68
    POSITIVE LOGITS
    唔
    0.81
    ⅽ
    0.81
    靣
    0.77
     Spouse
    0.76
     kanggo
    0.76
    Ꮩ
    0.72
     additionally
    0.72
    喺
    0.72
     adenine
    0.70
    פּ
    0.69
    Activations Density 0.376%

    No Known Activations