© 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. 31-GEMMASCOPE-2-RES-16K
    7. 6243
    Prev
    Next
    INDEX
    Explanations

    what contains or includes

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

    The neuron strongly activates on formal structural cue words—like “includes,” “including,” “discussion,” “provides,” “tells,” etc.—that introduce definitions, lists, or section content in technical or instructional text.

    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_31_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
     혹은
    0.82
     কিংবা
    0.71
    씩
    0.69
     ಅಥವಾ
    0.66
     illetve
    0.63
     či
    0.61
     nebo
    0.60
     cette
    0.59
     resulting
    0.59
     ili
    0.59
    POSITIVE LOGITS
    不仅
    1.97
     zarówno
    1.88
    不僅
    1.88
     sowohl
    1.84
    不仅仅
    1.84
     nejen
    1.77
     både
    1.73
    不但
    1.68
     both
    1.66
    both
    1.63
    Activations Density 0.630%

    No Known Activations