© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 10585
    Prev
    Next
    INDEX
    Explanations

    science and art

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

    This neuron fires on the first content‐bearing (nonfunction) word after a segment or paragraph break—that is, it highlights the opening “topic” word of each new block.

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

    words or phrases that appear near the middle or end of lengthy texts, particularly in contexts involving comparisons, alternatives, or situations with multiple items/people.

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

    comparative and quantitative language that marks degrees, counts, rankings, or alternative options in a description.

    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_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
     passé
    0.81
     Crater
    0.77
     Globalization
    0.77
     Method
    0.76
     grips
    0.74
    ని
    0.72
     Invitational
    0.71
     Coefficient
    0.71
     Couldn
    0.71
     -
    0.70
    POSITIVE LOGITS
     воздействия
    0.91
     дальней
    0.79
     получа
    0.78
    льня
    0.76
     последу
    0.75
     futuros
    0.75
     morgen
    0.75
    맑
    0.75
     осуществляется
    0.75
    に行く
    0.74
    Activations Density 0.000%

    No Known Activations