© 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-2-27B
    5. 34-GEMMASCOPE-RES-131K
    7. 30846
    Prev
    Next
    INDEX
    Explanations

    administrative division

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

    The neuron consistently lights up on formal technical tokens—like grammar nonterminal names (e.g. RANGE, INTEGER, ORDINAL, TOP), numeric literals (e.g. 1, 0), or other structured code‐oriented identifiers—so it’s detecting specialized symbols in programming or formal‐grammar contexts.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/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
     appartiennent
    -1.09
     gardent
    -0.90
     rozpoczę
    -0.86
    erunner
    -0.86
     przyjem
    -0.84
     commencent
    -0.83
     Broken
    -0.82
    thd
    -0.82
     produisent
    -0.81
     mężczyzn
    -0.80
    POSITIVE LOGITS
    ᄐ
    0.90
     opio
    0.83
     consequently
    0.83
     daggers
    0.82
    ểu
    0.81
    <0xD6>
    0.80
    ONOM
    0.79
    のか
    0.79
     psychos
    0.79
     subsequence
    0.78
    Activations Density 0.001%

    No Known Activations