© 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. 10-GEMMASCOPE-RES-131K
    7. 129705
    Prev
    Next
    INDEX
    Explanations

    District Court

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

    The neuron strongly activates on extremely common function words (like “is,” “was,” “the”) and other very high‐frequency tokens, effectively marking stopwords.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    common high‑frequency function words in general prose, especially auxiliary verbs, articles, and conjunctions.

    oai_token-act-pair · gpt-5Triggered by @jyhe0408

    italicized scientific or technical terms, particularly species names, gene names, and specialized terminology.

    oai_token-act-pair · claude-4-5-sonnetTriggered 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
     also
    -1.64
     pamamagitan
    -1.52
     both
    -1.34
     sierpnia
    -1.34
     barbati
    -1.30
     kwietnia
    -1.27
     even
    -1.22
     tanong
    -1.20
    delige
    -1.20
     folosit
    -1.18
    POSITIVE LOGITS
    would
    1.45
    ㇼ
    1.37
    matsu
    1.30
     entonces
    1.27
    他们
    1.27
     différences
    1.27
     ул
    1.26
    "-
    1.26
    .),
    1.23
    \\
    1.23
    Activations Density 0.033%

    No Known Activations