© 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. 128544
    Prev
    Next
    INDEX
    Explanations

    This neuron detects the main content words or topical keywords in a document—i.e. the informative, non-trivial terms that carry its subject matter.

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

    words followed by specific continuations

    np_acts-logits-general · gemini-2.5-flash-lite
    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
     будто
    -1.28
    んでる
    -1.26
     kvůli
    -1.19
     dingen
    -1.15
    移除
    -1.15
    结语
    -1.12
     junger
    -1.11
     “[
    -1.09
    んでいる
    -1.09
    Tapi
    -1.00
    POSITIVE LOGITS
     penerbangan
    1.30
     although
    1.25
     undangan
    1.22
    #
    1.20
     навіть
    1.18
     hitherto
    1.17
     kunjungan
    1.14
     Gemü
    1.13
     monies
    1.11
     very
    1.10
    Activations Density 0.045%

    No Known Activations