© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 63
    Prev
    Next
    INDEX
    Explanations

    JSON syntax and structure

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

    The neuron specializes in spotting numeric content—individual digits and multi‐digit number tokens (e.g. years, phone numbers, IDs, timestamps) often found in JSON or code.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    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
     and
    -1.55
     or
    -1.14
     for
    -1.09
    tubers
    -1.06
     However
    -1.05
     borealis
    -0.99
    voyez
    -0.98
    🤬
    -0.98
     During
    -0.97
     but
    -0.97
    POSITIVE LOGITS
     новом
    1.28
     tikai
    1.27
    킁
    1.23
     légèrement
    1.20
     cruelly
    1.20
     nieuwe
    1.17
     frequentemente
    1.16
     cilvē
    1.12
    mvh
    1.11
     kokona
    1.11
    Activations Density 0.043%

    No Known Activations