© 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. 10-GEMMASCOPE-RES-131K
    7. 7850
    Prev
    Next
    INDEX
    Explanations

    reasons and announcements

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

    The neuron activates on prominent header or section‐title words (typically uppercase or title‐cased) marking the start of a new section.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/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
     took
    -2.44
     extraordinarily
    -2.39
     during
    -2.36
    諴
    -2.31
     became
    -2.31
     takes
    -2.25
     went
    -2.20
     and
    -2.20
     unusually
    -2.20
     started
    -2.13
    POSITIVE LOGITS
      
    2.25
     REALLY
    2.22
    かっこ
    2.19
    だけでなく
    2.11
     dando
    2.11
               
    2.09
    Shouldn
    2.08
    referred
    2.05
          
    2.02
    だけではなく
    2.00
    Activations Density 0.002%

    No Known Activations