© 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. 22-GEMMASCOPE-RES-131K
    7. 39509
    Prev
    Next
    INDEX
    Explanations

    CJK languages, waste heat

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

    The neuron activates specifically on the acronym “CJK,” i.e. references to Chinese‐Japanese‐Korean languages.

    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
     undeniably
    -2.11
     cleverly
    -1.99
     epitom
    -1.96
    </h3>
    -1.87
    ↵
    -1.84
     encompasses
    -1.82
     Sadly
    -1.81
     Ironically
    -1.81
     могут
    -1.77
     exquisitely
    -1.77
    POSITIVE LOGITS
     REALLY
    2.02
    熟悉的
    1.98
     Brings
    1.95
    every
    1.87
    s
    1.83
     SPECIAL
    1.80
    to
    1.79
    受到了
    1.76
     PERFECT
    1.75
     crappy
    1.70
    Activations Density 0.000%

    No Known Activations