© 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
    3. Gemma-2-27B
    5. 34-GEMMASCOPE-RES-131K
    7. 78289
    Prev
    Next
    INDEX
    Explanations

    punctuation and specific titles

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

    The neuron strongly activates on formal legal-opinion language—especially past-tense procedural verbs and markers (e.g. “compiled,” “printed;,” “opinion,” “counsel,” “We”) typical of court‐opinion texts.

    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
     thus
    -0.87
     via
    -0.85
     hence
    -0.84
     aforesaid
    -0.84
    Инг
    -0.79
    ďaka
    -0.79
     thence
    -0.75
     like
    -0.75
    Thus
    -0.73
     administração
    -0.72
    POSITIVE LOGITS
     then
    0.85
     اگر
    0.85
     عرب
    0.82
    anadium
    0.82
    ถ้า
    0.80
    ੋ
    0.80
    附
    0.79
     если
    0.79
     اله
    0.77
    </tr>
    0.77
    Activations Density 0.005%

    No Known Activations