© 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. 34-GEMMASCOPE-RES-131K
    7. 4227
    Prev
    Next
    INDEX
    Explanations

    clauses starting with "When"

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

    The neuron strongly fires on the very first content word of a new sentence or instruction step (e.g. “When,” “I,” “But,” “uploading,” “process,” “notice,” “fiber,” etc.), i.e. it detects sentence- or clause-initial tokens.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_34/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
     also
    -0.94
     МЕ
    -0.94
    我已经
    -0.83
     Empieza
    -0.81
     nô
    -0.80
     jambe
    -0.80
     lokalen
    -0.79
    IRY
    -0.78
    PEP
    -0.77
     Mô
    -0.77
    POSITIVE LOGITS
     saw
    1.02
     vừa
    0.94
    提到
    0.93
     зустрі
    0.91
     sees
    0.91
     see
    0.84
     was
    0.82
     comes
    0.80
     heard
    0.79
     approaching
    0.79
    Activations Density 0.006%

    No Known Activations