© 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. 67484
    Prev
    Next
    INDEX
    Explanations

    class implementing

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

    The neuron strongly activates on technical two‐word phrases where a modifier (like an adjective or preposition) pairs with a noun—e.g. “concrete class,” “extended interface,” “in vivo,” “ideal number,” “corn fields,” “another list.”

    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
    Läh
    -1.08
    ể
    -1.02
     emplear
    -1.02
    ſte
    -1.02
     hornear
    -0.97
     fiabilidad
    -0.96
     první
    -0.95
    vgili
    -0.95
    ėtų
    -0.94
    wym
    -0.93
    POSITIVE LOGITS
     also
    1.34
     still
    0.99
     normaux
    0.95
    hlander
    0.92
     universo
    0.92
     even
    0.92
    。
    0.92
    还是要
    0.91
     faudrait
    0.90
     difer
    0.90
    Activations Density 0.085%

    No Known Activations