© 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
    3. Gemma-2-27B
    5. 10-GEMMASCOPE-RES-131K
    7. 1726
    Prev
    Next
    INDEX
    Explanations

    scholarly and tainted

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

    The neuron predominantly fires on longer, multi‐syllable content words (e.g. “scholarly,” “tainted,” “sstream”), effectively picking out uncommon or specialized vocabulary.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     The
    -3.23
     Then
    -2.72
    莛
    -2.38
     enorme
    -2.23
    僦
    -2.17
     avec
    -2.11
     [
    -2.11
     coloca
    -2.09
     venden
    -2.09
     imma
    -2.08
    POSITIVE LOGITS
    *
    3.11
    ly
    2.84
    罨
    2.58
    When
    2.55
    ”？
    2.53
    that
    2.53
    欵
    2.45
    the
    2.44
    ｛
    2.44
    Despite
    2.41
    Activations Density 0.004%

    No Known Activations