© 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
    2. Gemma-2-27B
    3. 34-GEMMASCOPE-RES-131K
    4. 58373
    Prev
    Next
    INDEX
    Explanations

    version number format

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

    The neuron fires on structural or “section‐header” tokens—things like Ruby block delimiters (do and the pipe “|”), XML/JSON field names (Version, systemVersion), license section keywords (person, conditions), or newswire labels (ElectronicallyFiled, Marketwire)—i.e. markers that signal the start of a new code block or document section.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/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
     others
    -0.73
    LAS
    -0.68
    estan
    -0.66
     ​
    -0.63
     mog
    -0.63
    〚
    -0.62
     Eigenschaften
    -0.62
    implementation
    -0.61
     attributes
    -0.61
     シリーズ
    -0.61
    POSITIVE LOGITS
    Lane
    0.75
     mitu
    0.74
    ☲
    0.74
    Chester
    0.71
     Lane
    0.69
     tils
    0.69
     TER
    0.69
     Aktiv
    0.68
    lámpara
    0.67
    шься
    0.67
    Activations Density 0.065%

    No Known Activations