© 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
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 75139
    Prev
    Next
    INDEX
    Explanations

    code, legal, or concluding phrases

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

    The neuron consistently fires on tokens that mark the end of a block or section—such as “done,” “},” “EOS,” standalone periods or headings like “Defendant”—i.e. document delimiters or section-boundary markers.

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

    the end of code blocks or statements in programming languages.

    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_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
    啲
    -0.98
     sowohl
    -0.97
    これは
    -0.96
     yetu
    -0.93
    や
    -0.91
     includ
    -0.90
     lect
    -0.90
    न्छ
    -0.90
    gulama
    -0.90
    Ǟ
    -0.90
    POSITIVE LOGITS
     vždy
    0.98
    就可以了
    0.93
    さんへ
    0.92
    права
    0.91
    に追加
    0.91
    )}$.
    0.91
    друг
    0.89
     isSuccess
    0.89
    様に
    0.88
     имеет
    0.88
    Activations Density 0.001%

    No Known Activations