© 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
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 84377
    Prev
    Next
    INDEX
    Explanations

    sequences leading to a result

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

    The neuron fires most strongly on discourse-level transition markers—words like “and,” “then,” “if,” or sentence-initial pronouns (e.g. “She,” “He,” “This”) that introduce new clauses or narrative turns.

    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
    
    -1.16
    Parce
    -1.06
    Пока
    -1.03
    Semua
    -1.03
    Seems
    -1.02
    Selon
    -1.00
    Если
    -0.99
    Buat
    -0.99
    桖
    -0.97
    Terima
    -0.95
    POSITIVE LOGITS
     so
    2.00
    于是
    1.87
     decided
    1.55
     therefore
    1.49
     took
    1.47
     went
    1.35
    於是
    1.31
    จึง
    1.30
     поэтому
    1.29
     hence
    1.28
    Activations Density 0.036%

    No Known Activations