© 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
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 66163
    Prev
    Next
    INDEX
    Explanations

    foreign lowercase words

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

    The neuron is highly sensitive to structural punctuation characters—especially colons, commas, slashes, and the like—used as field or label separators in formatted text.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_22/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
    してみると
    -1.09
    文獻
    -1.08
     ẞ
    -1.08
     regn
    -1.08
     swiftly
    -1.04
     wali
    -0.96
     Ename
    -0.95
     boya
    -0.94
    ‌تر
    -0.94
     these
    -0.94
    POSITIVE LOGITS
     saliendo
    1.38
     </
    1.18
     trám
    1.12
    蹭
    1.11
    $};
    1.11
    弱い
    1.10
     →
    1.10
     lú
    1.09
    bný
    1.07
     podró
    1.05
    Activations Density 0.000%

    No Known Activations