© 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. 22-GEMMASCOPE-RES-131K
    7. 66932
    Prev
    Next
    INDEX
    Explanations

    component tags and attributes

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

    This neuron fires on the very first content word of each text segment—i.e. it detects the initial “topic” token that starts a new section or question.

    oai_token-act-pair · o4-miniTriggered 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
     óra
    -0.84
    deras
    -0.82
     his
    -0.80
    是否
    -0.80
    ouder
    -0.79
     között
    -0.78
     jego
    -0.77
     member
    -0.76
    lohy
    -0.75
    信息
    -0.75
    POSITIVE LOGITS
    üyada
    0.93
     Sebelum
    0.87
    çando
    0.86
     uj
    0.85
     Apalagi
    0.82
     Tuttavia
    0.82
     Inicial
    0.80
     expensive
    0.79
    vasive
    0.79
     actionMode
    0.79
    Activations Density 0.092%

    No Known Activations