© 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. 39746
    Prev
    Next
    INDEX
    Explanations

    so much

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

    The neuron fires on tokens that serve as discourse‐level connectors or markers—things like conjunctions (and, so), reporting verbs (discussed), possessive ’s, hyphens, titles/speaker labels (Mr, MAN)—i.e. words and symbols that link clauses or indicate shifts in speaker or structure.

    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
    也太
    -0.85
     two
    -0.82
     three
    -0.81
     Ta
    -0.79
     Newswire
    -0.78
     especie
    -0.77
    有多少
    -0.77
     manipulation
    -0.77
    Archiv
    -0.76
     Rückseite
    -0.76
    POSITIVE LOGITS
     svoju
    0.99
     váš
    0.98
     ešte
    0.94
    现在
    0.88
    raria
    0.85
    ㉖
    0.85
    英语
    0.84
    loem
    0.84
    στή
    0.83
     énorme
    0.83
    Activations Density 0.007%

    No Known Activations