© 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. 14147
    Prev
    Next
    INDEX
    Explanations

    words ending in 'ions', 'ed', or 'ure'

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

    The neuron consistently lights up on internal subword fragments (BPE tokens) that are parts of longer, often technical or complex words (e.g. “effic,” “emi,” “ions,” “device,” “manner”).

    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
    Kenapa
    -1.37
    Tentang
    -1.27
    Hanya
    -1.20
    Before
    -1.20
    Several
    -1.19
    Another
    -1.19
    Приятного
    -1.19
    gö
    -1.16
    Jangan
    -1.09
    Some
    -1.08
    POSITIVE LOGITS
    ᯾
    1.34
    或
    1.25
     others
    1.24
     gita
    1.24
    TUAL
    1.21
    itic
    1.21
    poptotic
    1.20
    。
    1.18
     then
    1.16
     &.
    1.13
    Activations Density 0.004%

    No Known Activations