© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 25080
    Prev
    Next
    INDEX
    Explanations

    image reconstruction

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

    The neuron strongly activates on long, multi‐part or compound tokens—especially technical or biomedical terms (e.g. “fullWord,” “reconstruct,” “autosomal,” “otitis,” “media”).

    oai_token-act-pair · o4-mini
    Triggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Negative Logits
    lendir
    -0.93
    aternion
    -0.93
     žád
    -0.91
     純正
    -0.88
    玩意
    -0.86
    ꪹ
    -0.85
    vector
    -0.85
    ÄT
    -0.85
    вши
    -0.84
    крыть
    -0.84
    POSITIVE LOGITS
     that
    0.97
     what
    0.91
     there
    0.90
    ोड़
    0.81
    шений
    0.78
     expérimentés
    0.77
     same
    0.77
    ਊ
    0.77
    летия
    0.77
     if
    0.75
    Activations Density 0.067%

    No Known Activations