© 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. 34-GEMMASCOPE-RES-131K
    7. 31543
    Prev
    Next
    INDEX
    Explanations

    Case in point

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

    The neuron fires strongly on programming‐style symbols and technical metadata tokens—things like “std”, “::”, uppercase field names (e.g. “Content-Type”, “User-Agent”), version numbers, and similar code or header identifiers.

    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
     cold
    -0.89
    เต็ม
    -0.76
     Dokumentation
    -0.71
     útiles
    -0.70
     холод
    -0.70
     documentation
    -0.69
     phí
    -0.68
     مُعرِّف
    -0.68
     fluorine
    -0.68
    chées
    -0.67
    POSITIVE LOGITS
    wet
    0.75
    breakable
    0.71
     Greenberg
    0.69
    猛地
    0.68
    area
    0.68
    Advisor
    0.67
     moz
    0.66
    bros
    0.65
     Fucking
    0.65
    voort
    0.65
    Activations Density 0.066%

    No Known Activations