© 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. 22-GEMMASCOPE-RES-131K
    7. 15678
    Prev
    Next
    INDEX
    Explanations

    code syntax and programming constructs

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

    This neuron responds to lines or tokens that are unusually heavy in punctuation or symbol characters—e.g. long runs of “=”, code‐block braces/keywords, or other punctuation-dense separators.

    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
     publico
    -1.06
     menjawab
    -1.04
     adorno
    -1.02
     both
    -0.96
    šanai
    -0.95
     only
    -0.94
    hörige
    -0.94
    İŞ
    -0.93
    nextToken
    -0.92
     monstruo
    -0.91
    POSITIVE LOGITS
     '',
    0.96
     alcuni
    0.96
     teilen
    0.94
    whom
    0.91
     crescere
    0.89
     terutama
    0.88
     prospet
    0.88
     enriquec
    0.88
    richtet
    0.87
    ⪢
    0.87
    Activations Density 0.004%

    No Known Activations