© 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. 84312
    Prev
    Next
    INDEX
    Explanations

    Programming language keywords and symbols

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

    The neuron strongly activates on programming‐style tokens—e.g. identifiers (String, class, test), symbol prefixes ($, _), keywords (function), and other code punctuation—so it’s essentially detecting code fragments.

    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
     apartament
    -1.31
     pousada
    -1.21
    があり
    -1.18
    ungalow
    -1.12
    libat
    -1.09
    參考資料
    -1.05
    例文
    -1.04
    醍
    -1.04
    obatan
    -1.04
    があります
    -1.04
    POSITIVE LOGITS
    際には
    1.58
    際に
    1.55
    際は
    1.42
    ごめん
    1.40
    あぁ
    1.40
    さあ
    1.36
    さぁ
    1.30
    総評
    1.28
    来週
    1.28
    えっ
    1.25
    Activations Density 0.047%

    No Known Activations