© 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. 30776
    Prev
    Next
    INDEX
    Explanations

    names and dialogue tags

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

    This neuron detects document-structural labels and speaker‐attribution markers—e.g. chapter headings (“CHAPTER I”), quotation marks around dialogue, and attribution words like “said” alongside character names.

    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
     shit
    -1.08
     bloody
    -1.04
     damn
    -1.01
     hombre
    -0.94
    っており
    -0.93
     vợ
    -0.91
    yz
    -0.89
    itás
    -0.88
    Bloody
    -0.87
     but
    -0.86
    POSITIVE LOGITS
     педагоги
    1.02
    少女
    1.01
     girls
    1.01
    poň
    0.91
     herself
    0.89
     envia
    0.85
    cupa
    0.84
     afghan
    0.84
     Economies
    0.84
    Bha
    0.84
    Activations Density 0.011%

    No Known Activations