© 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. 34-GEMMASCOPE-RES-131K
    7. 108630
    Prev
    Next
    INDEX
    Explanations

    figures and supporting information

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

    The neuron fires on “preamble” and metadata tokens—things like LaTeX commands and package names, section‐ or license‐related words (e.g. “\usepackage”, “for”, “its”, “legal”, “reference”)—marking up document structure rather than natural text.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/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
    threads
    -0.84
    turned
    -0.82
     imprime
    -0.81
    preferably
    -0.81
     egyik
    -0.79
    jaya
    -0.79
    ギュラー
    -0.77
     lagen
    -0.76
     geringe
    -0.76
    drugs
    -0.76
    POSITIVE LOGITS
     figures
    1.30
    figures
    1.15
     figs
    1.13
     Figures
    1.06
    Figures
    1.05
     figure
    0.98
     figuras
    0.97
     tables
    0.92
     Figs
    0.91
    idity
    0.85
    Activations Density 0.004%

    No Known Activations