© 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. 10-GEMMASCOPE-RES-131K
    7. 113338
    Prev
    Next
    INDEX
    Explanations

    change and transformation

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

    This neuron fires most strongly on unusually long or rare subword tokens—that is, it detects long, low-frequency words (broken into multiple subword pieces).

    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
    "
    -2.78
    (
    -2.11
    1
    -2.08
    та
    -2.06
    >
    -2.03
    IBLIO
    -1.94
    -
    -1.92
    л
    -1.91
    when
    -1.90
    скому
    -1.90
    POSITIVE LOGITS
    ↵
    3.02
     Aufenthalt
    2.34
    鷄
    2.31
    ↀ
    2.30
    葑
    2.22
     Ausblick
    2.22
     Preisen
    2.11
     Gebiete
    2.05
     (“
    1.99
     másik
    1.99
    Activations Density 0.005%

    No Known Activations