© 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. 22-GEMMASCOPE-RES-131K
    7. 95841
    Prev
    Next
    INDEX
    Explanations

    keywords and identifiers

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

    The neuron strongly activates on isolated technical identifiers or jargon-like tokens (e.g. Android, commercial, license, help, open, cart) rather than on common words or syntax.

    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
    cresc
    -1.30
    Skal
    -1.18
    amateur
    -1.16
    enfer
    -1.16
     говорится
    -1.16
    aigu
    -1.14
    amplific
    -1.13
    roland
    -1.13
    Hilsen
    -1.11
    Antwoord
    -1.10
    POSITIVE LOGITS
     as
    1.59
    Info
    1.30
     information
    1.29
     does
    1.24
    💩
    1.21
     versión
    1.16
    U
    1.15
    info
    1.14
    信息
    1.14
    Data
    1.11
    Activations Density 0.021%

    No Known Activations