© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-2-27B
    3. 22-GEMMASCOPE-RES-131K
    4. 15199
    Prev
    Next
    INDEX
    Explanations

    geographical names and numbers

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

    The neuron specializes in picking out structured factual tokens—especially numbers (dates, counts, measurements), acronyms, and proper‐name terms that typically appear in metadata or info‐box–style descriptions.

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    location names (cities, states, countries, and venues) in formal text about events or places.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    capitalized named entities and formal encyclopedic metadata, especially places, dates, organizations, and event or competition details.

    oai_token-act-pair · gpt-5Triggered 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
     that
    -1.70
     from
    -1.47
     then
    -1.45
    :
    -1.41
     to
    -1.39
     for
    -1.36
     so
    -1.34
     because
    -1.31
     of
    -1.28
     by
    -1.27
    POSITIVE LOGITS
     conster
    1.41
     interrom
    1.24
     interro
    1.22
     corrom
    1.21
     aldrig
    1.20
    罍
    1.20
     negativa
    1.18
     filmi
    1.17
     alcool
    1.16
     éta
    1.16
    Activations Density 0.328%

    No Known Activations