© 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. 13767
    Prev
    Next
    INDEX
    Explanations

    abstract concepts and virtues

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

    The neuron strongly activates on single, weighty abstract or philosophical nouns—words like “truth,” “beauty,” “sin,” “death,” “illusion,” etc.—marking major existential or moral concepts.

    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
    Almost
    -1.23
    Each
    -1.20
    Start
    -1.20
    leggen
    -1.13
     for
    -1.10
     dadas
    -1.09
    וּ
    -1.08
     inggris
    -1.06
    About
    -1.05
    Get
    -1.04
    POSITIVE LOGITS
     itself
    1.57
     deceit
    1.14
     its
    1.13
     prevenção
    1.08
    ifr
    1.06
    があれば
    1.05
    ufc
    1.05
    その
    1.05
    かもしれませんが
    1.05
     propriedades
    1.05
    Activations Density 0.019%

    No Known Activations