© 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. 6357
    Prev
    Next
    INDEX
    Explanations

    for illustrative purposes

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

    The neuron consistently fires on the main noun or headword that completes a fixed multiword phrase (e.g. the “order” in “back order,” “sitting” in “sitting position,” “effect” in “take effect,” etc.).

    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
     които
    0.69
     която
    0.69
     който
    0.67
     który
    0.66
     зокрема
    0.64
     która
    0.63
     kterou
    0.61
     którą
    0.61
     která
    0.61
     które
    0.60
    POSITIVE LOGITS
    នូវ
    0.73
    만을
    0.66
     presso
    0.61
    อะไร
    0.60
     alumina
    0.60
    alka
    0.59
     alcoved
    0.58
     óta
    0.58
     modulo
    0.57
     oleh
    0.57
    Activations Density 0.212%

    No Known Activations