© 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. 117500
    Prev
    Next
    INDEX
    Explanations

    version numbers/* or specific clauses

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

    The neuron is not picking out any particular semantic category but rather fires strongly on tokens that occur many times in a snippet (e.g. repeated “file” in imports), i.e. it’s acting like a term-frequency detector.

    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
     tress
    -0.74
    被称为
    -0.73
     spandex
    -0.71
     discer
    -0.70
     mosquitos
    -0.70
     matt
    -0.69
     Kokos
    -0.69
     milliseconds
    -0.68
     catastrophes
    -0.68
    loy
    -0.67
    POSITIVE LOGITS
     orice
    0.72
    minar
    0.71
    ும்
    0.70
     Sampling
    0.68
     Flugzeug
    0.68
    Ḫ
    0.68
    ticelli
    0.67
    PackageInfo
    0.67
    <0x9C>
    0.67
     sys
    0.65
    Activations Density 0.041%

    No Known Activations