© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-2-27B
    3. 10-GEMMASCOPE-RES-131K
    4. 99075
    Prev
    Next
    INDEX
    Explanations

    linking words and phrases

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

    The neuron selectively activates on high-frequency function words—especially conjunctions and similar connectors (e.g. and, but, or, however, who, depending)—that link clauses.

    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
    ▰
    -1.58
    釕
    -1.52
    も多く
    -1.52
    唯一的
    -1.49
     nurtured
    -1.48
     shunned
    -1.48
    1
    -1.46
     thwarted
    -1.46
    悄悄
    -1.45
     predominantly
    -1.41
    POSITIVE LOGITS
     as
    1.85
    ↵
    1.57
     and
    1.55
     of
    1.52
     from
    1.49
     had
    1.45
     but
    1.41
    </tr>
    1.36
    quelize
    1.33
     (
    1.30
    Activations Density 0.013%

    No Known Activations