© 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. 22-GEMMASCOPE-RES-131K
    7. 129438
    Prev
    Next
    INDEX
    Explanations

    The neuron mainly picks up isolated numeric tokens used as version or section labels (e.g. standalone “0” in a “0.5.2” version or a solitary “1” in a heading).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    medical and social issues

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

    line numbers or version numbers at the beginning of a document or section.

    oai_token-act-pair · claude-3-7-sonnet-20250219Triggered by @neilrathi
    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
     both
    -2.38
     all
    -2.34
     several
    -2.28
     more
    -2.25
     this
    -2.25
     three
    -2.25
     because
    -2.17
     two
    -2.13
     those
    -2.13
     four
    -2.09
    POSITIVE LOGITS
     delightfully
    2.13
     REALLY
    2.09
     unbelievably
    2.03
     amazingly
    1.99
     trám
    1.99
     insanely
    1.92
     sooo
    1.91
     soooo
    1.91
     ส์
    1.90
     captivating
    1.89
    Activations Density 0.000%

    No Known Activations