© 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-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 457
    Prev
    Next
    INDEX
    Explanations

    d... words like drought, dune, dating, decline

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

    The neuron is highly activated by tokens that begin a word with “D” (e.g. D-, De-, Dr-, SD-, or d- at the start of a token).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-pt/resid_post/layer_31_width_16k_l0_medium
    Prompts (Dashboard)
    392,802 prompts, 256 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     seeing
    0.36
     heure
    0.33
     airtight
    0.32
     rang
    0.32
     hops
    0.32
     tenor
    0.31
     mah
    0.31
     strong
    0.30
     seafront
    0.30
     coaster
    0.30
    POSITIVE LOGITS
    ifferenti
    0.55
    ifferentiated
    0.54
    ifferentiate
    0.54
    ocument
    0.52
    ifferent
    0.51
    ifferentiating
    0.51
    velopment
    0.50
    irection
    0.50
    ressed
    0.48
    ifference
    0.48
    Activations Density 0.367%

    No Known Activations