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

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 22-GEMMASCOPE-RES-131K
    7. 35509
    Prev
    Next
    INDEX
    Explanations

    sentence start words

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

    The neuron is highly sensitive to sentence- or paragraph-initial tokens (especially capitalized words like “According,” “The,” etc.), effectively marking the start of new sentences.

    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
     рассказывает
    -1.08
    ということで
    -1.07
     there
    -1.05
    erobic
    -1.02
     various
    -1.02
    کتری
    -1.00
    スプレー
    -1.00
     voici
    -0.99
    結果は
    -0.98
    <bos>
    -0.98
    POSITIVE LOGITS
     Когда
    1.26
     If
    1.26
    ditionally
    1.21
    ientras
    1.15
     quando
    1.13
    ₱
    1.13
     gdyby
    1.13
    筈
    1.13
    事です
    1.13
    Although
    1.10
    Activations Density 0.029%

    No Known Activations