© 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
    3. Gemma-2-27B
    5. 10-GEMMASCOPE-RES-131K
    7. 104558
    Prev
    Next
    INDEX
    Explanations

    counts and percentages

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

    The neuron strongly activates on the hyphen character when it appears in numeric expressions (e.g. “Fifty-two,” “Thirty-two”), effectively spotting numbers written with dashes.

    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
    0
    -2.61
    a
    -2.31
    ↵
    -2.16
    퐸
    -2.13
     informée
    -2.13
    亖
    -1.91
    s
    -1.89
    état
    -1.88
     også
    -1.77
    可以说
    -1.76
    POSITIVE LOGITS
     ignite
    2.30
    
    2.25
    <bos>
    2.02
    --;
    1.95
    ]-->
    1.88
    隄
    1.86
    ).
    1.85
    趵
    1.85
     }));
    1.84
    :],
    1.84
    Activations Density 0.013%

    No Known Activations