© 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
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 75361
    Prev
    Next
    INDEX
    Explanations

    code and programming keywords

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

    The neuron strongly activates on tiny, often single‐character code tokens (e.g. one‐letter variables like “c”, “k”, “d”, the “*” operator, etc.) that commonly appear in heavily minified or obfuscated code.

    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
     Baş
    -1.13
    recycled
    -1.03
     croit
    -1.03
     چهارم
    -1.02
    ⸒
    -1.01
     tersenyum
    -1.00
     at
    -0.99
    glichen
    -0.98
     placée
    -0.96
     supermercado
    -0.96
    POSITIVE LOGITS
    has
    1.03
     метода
    0.99
    蹦
    0.97
    i
    0.96
    one
    0.96
    u
    0.95
    
    0.94
    END
    0.93
    けど
    0.93
    needs
    0.93
    Activations Density 0.001%

    No Known Activations