© 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. 24-GEMMASCOPE-2-RES-16K
    7. 10849
    Prev
    Next
    INDEX
    Explanations

    markdown files and code structures

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

    This neuron doesn’t respond to ordinary running text but spikes on the isolated “metadata” or section‐marking tokens that signal shifts in document structure (things like headings, legal/disclaimer markers, season indicators, URLs, and other genre‐signaling words).

    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_24_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
     difund
    0.89
     ciudades
    0.85
     Jacqueline
    0.80
     Cities
    0.76
     empiezan
    0.75
    膈
    0.74
     extol
    0.74
    ovich
    0.72
     traducir
    0.72
     Christina
    0.72
    POSITIVE LOGITS
    к
    0.89
    Го
    0.81
    d
    0.79
    ри
    0.79
    بي
    0.76
    גר
    0.75
    GU
    0.73
    ת
    0.73
     зо
    0.73
    ла
    0.71
    Activations Density 0.000%

    No Known Activations