© 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
    3. Gemma-2-27B
    5. 10-GEMMASCOPE-RES-131K
    7. 112021
    Prev
    Next
    INDEX
    Explanations

    can + do / obtain / use

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

    The neuron fires on modal‐auxiliary and action‐suggestion words (like “can,” “could,” “use,” “would”) that express ability, possibility, or recommendation.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    ໊
    -1.95
     Père
    -1.81
    ܖ
    -1.80
    ܧ
    -1.77
    𐐀
    -1.73
     journaliste
    -1.67
    ܤ
    -1.62
    ڧ
    -1.58
    ݯ
    -1.57
    錀
    -1.57
    POSITIVE LOGITS
     from
    2.38
     that
    2.22
     acquire
    1.87
     which
    1.85
     improve
    1.83
     with
    1.82
     obtain
    1.69
     sapore
    1.62
     lower
    1.62
    us
    1.59
    Activations Density 0.065%

    No Known Activations