© 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. 104358
    Prev
    Next
    INDEX
    Explanations

    how, what, where, why, who

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

    This neuron detects clause-connecting tokens—chiefly commas and coordinating/subordinating conjunctions like “and,” “but,” “if,” “when,” and “how.”

    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
    居る
    -1.01
    
    -0.98
     andre
    -0.98
     других
    -0.96
    GRAFÍA
    -0.95
     unserem
    -0.94
     سایر
    -0.94
    鯪
    -0.94
    GRAFIA
    -0.93
    ceptar
    -0.93
    POSITIVE LOGITS
     how
    2.86
     what
    2.19
     who
    1.63
     Who
    1.38
     How
    1.35
     where
    1.31
     whether
    1.30
    how
    1.23
     why
    1.14
     hvem
    1.13
    Activations Density 0.051%

    No Known Activations