© 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. 25429
    Prev
    Next
    INDEX
    Explanations

    programming keywords and methods

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

    The neuron responds to programming code patterns—specifically recognizing function or method call tokens (e.g. “require”, “Split”, “parse”) rather than ordinary 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
     need
    -1.12
     others
    -1.05
    を
    -1.04
    cip
    -0.99
    ulose
    -0.94
    aire
    -0.92
     metodi
    -0.91
    
    -0.89
     olacak
    -0.89
     further
    -0.89
    POSITIVE LOGITS
     from
    1.17
    有名な
    0.98
    FFEE
    0.94
     alkot
    0.93
    所说的
    0.91
    سمبر
    0.91
     újra
    0.91
    itás
    0.89
     Finger
    0.89
    どのような
    0.88
    Activations Density 0.030%

    No Known Activations