© 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. 5652
    Prev
    Next
    INDEX
    Explanations

    unique identifier

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

    This neuron strongly activates on tokens that begin new sentences or segments, especially the first (often capitalized) word at the start of a sentence.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    sentence or section beginnings, particularly in formal or structured text.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    capitalized, high-salience tokens at the starts of sentences or titles (including proper names, interjections, and headline-style words).

    oai_token-act-pair · gpt-5Triggered 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
     คน
    0.62
     greats
    0.60
    উত্তর
    0.57
     agonist
    0.57
     wills
    0.56
     signings
    0.55
     celebs
    0.55
     ones
    0.54
     monarchy
    0.54
     terapi
    0.54
    POSITIVE LOGITS
    oretically
    1.49
    ostante
    1.28
    theless
    1.08
    смотря
    1.05
    spapers
    0.94
    spaper
    0.94
    mittedly
    0.93
    скольку
    0.93
    withstanding
    0.91
    swering
    0.90
    Activations Density 8.878%

    No Known Activations