© 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. 10-GEMMASCOPE-RES-131K
    7. 68159
    Prev
    Next
    INDEX
    Explanations

    starts with Leb or Marlene or is wrought

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

    Explanation of neuron 4 behavior: the main thing this neuron does is find subword tokens corresponding to proper names (capitalized name fragments).

    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
    潆
    -3.30
     savent
    -2.91
    氿
    -2.89
     veulent
    -2.86
    狴
    -2.84
    湏
    -2.84
    蟥
    -2.78
    鈇
    -2.73
    樒
    -2.72
     voient
    -2.69
    POSITIVE LOGITS
    an
    2.67
    I
    2.52
    ﾄ
    2.47
    悒
    2.45
     With
    2.28
    ."
    2.27
    ffeine
    2.20
     незавер
    2.19
    In
    2.17
     five
    2.17
    Activations Density 0.005%

    No Known Activations