© 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. 66722
    Prev
    Next
    INDEX
    Explanations

    whomever or whoever holds position

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

    The neuron is keyed to surface tokens carrying concrete, content‐bearing information—especially numbers (years, dates, counts) and proper names (organizations, venues, teams).

    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
     quando
    -1.03
    ственном
    -1.01
    тельными
    -0.96
     incluso
    -0.96
    sometimes
    -0.95
    même
    -0.95
     lorsque
    -0.95
     जहां
    -0.93
    EVEN
    -0.92
     latine
    -0.92
    POSITIVE LOGITS
     whichever
    1.93
     either
    1.91
     whoever
    1.80
     whome
    1.63
     someone
    1.55
     any
    1.55
    hichever
    1.47
     preferably
    1.46
     либо
    1.40
    Either
    1.38
    Activations Density 0.071%

    No Known Activations