© 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. 104685
    Prev
    Next
    INDEX
    Explanations

    lot of, few know, since Java

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

    verb forms and phrases indicating concession, contrast, or comparison (such as "though", "as", "be", "were", "of those").

    oai_token-act-pair · claude-4-5-sonnetTriggered by @bfgcn7frf5

    The neuron is a “content‐word” detector – it lights up on semantically meaningful words (nouns, verbs, adjectives) rather than on purely functional tokens.

    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
     gitar
    -1.70
     kompres
    -1.66
     silikon
    -1.59
     kompakt
    -1.58
     mikrofon
    -1.55
     elektron
    -1.54
     rezerv
    -1.54
    eeeeeeee
    -1.53
     aktiv
    -1.48
     produkt
    -1.48
    POSITIVE LOGITS
     after
    1.82
     do
    1.72
     –
    1.70
     at
    1.65
    h
    1.60
     is
    1.58
     both
    1.52
     beispielsweise
    1.44
    -
    1.38
    )–
    1.38
    Activations Density 0.141%

    No Known Activations