© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 127943
    Prev
    Next
    INDEX
    Explanations

    The neuron strongly activates on pure numeric tokens (i.e. multi‐digit numbers or IDs) rather than on letters or punctuation.

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

    Is questions

    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
     housse
    -1.06
     parque
    -1.04
     fábricas
    -1.02
    edade
    -1.02
     yelek
    -1.01
     selva
    -1.01
     tebal
    -1.00
     piscinas
    -1.00
     ücretsiz
    -0.98
     寿
    -0.98
    POSITIVE LOGITS
    ↵↵
    1.03
    全都
    0.99
     früheren
    0.98
     alcune
    0.97
     membro
    0.96
     struts
    0.95
     REALLY
    0.94
    3
    0.94
     iedereen
    0.94
     latein
    0.94
    Activations Density 0.007%

    No Known Activations