© 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. 11239
    Prev
    Next
    INDEX
    Explanations

    aggressive interactions

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

    The neuron fires strongly on high-information “headline” tokens—the key domain or topic words (e.g. “method,” “bankruptcy,” “Pepsi’s,” “I”) that kick off a new section or sentence and signal the subject matter.

    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
    en
    0.94
    e
    0.91
    m
    0.91
    z
    0.86
    in
    0.84
    h
    0.82
     трех
    0.81
    es
    0.80
    i
    0.79
     четырех
    0.79
    POSITIVE LOGITS
     blatant
    0.97
    是没有
    0.97
    這些
    0.82
     sesu
    0.80
    的這個
    0.80
    ಿದ್ದರೆ
    0.80
    ইহার
    0.79
    这些
    0.79
     ignorant
    0.79
     blatantly
    0.79
    Activations Density 0.000%

    No Known Activations