© 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. 64557
    Prev
    Next
    INDEX
    Explanations

    precedential, follow, despite, alias, Q

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

    The neuron is keyed to discourse-structuring tokens—sentence-initial or “transition” words and phrases (e.g. “There,” “Follow,” “e.g.,” “de Tocqueville,” “simplifications”) that mark the flow of an argument.

    oai_token-act-pair · o4-miniTriggered 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
    Not in Any Lists

    No Comments

    Negative Logits
     underdog
    -0.90
     cuire
    -0.88
    ティブ
    -0.87
     наси
    -0.81
     funcionario
    -0.81
    하면
    -0.80
    Xbox
    -0.79
    гиоз
    -0.79
    铵
    -0.79
     就是
    -0.77
    POSITIVE LOGITS
    PreInfinity
    0.93
    clero
    0.91
     Flüs
    0.88
    ivano
    0.88
     ويك
    0.87
    Melhor
    0.87
     cuyo
    0.87
    碉
    0.86
     need
    0.84
    ticed
    0.84
    Activations Density 0.000%

    No Known Activations