© 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. 10-GEMMASCOPE-RES-131K
    4. 124016
    Prev
    Next
    INDEX
    Explanations

    titles, relationships, and specific entities

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

    The neuron fires on the headings or labels that introduce numbered list entries (e.g. “Exercise,” “Diocese,” “Provincial,” “Piano,” “Emperor”)—i.e. words immediately preceding numeric identifiers in a structured list.

    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
     
    -1.95
     réve
    -1.66
     then
    -1.63
     itinéraire
    -1.55
     métro
    -1.55
     éta
    -1.53
     rellen
    -1.52
    鹩
    -1.50
    。)
    -1.48
    /",
    -1.47
    POSITIVE LOGITS
     dará
    1.70
     четырех
    1.55
     from
    1.55
     during
    1.54
     without
    1.52
     назвать
    1.51
     on
    1.48
    бил
    1.47
     to
    1.45
     realizará
    1.43
    Activations Density 0.039%

    No Known Activations