© 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
    3. Gemma-3-12B
    5. 24-GEMMASCOPE-2-RES-16K
    7. 1319
    Prev
    Next
    INDEX
    Explanations

    The neuron is most strongly activated by high‐frequency function tokens—common punctuation (e.g. commas) and very frequent small words like “more,” “in,” etc.

    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_24_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
    idades
    0.39
     secretive
    0.39
    DQ
    0.39
    NING
    0.38
    ר
    0.38
    े
    0.38
    פ
    0.38
    गें
    0.38
    뵐
    0.37
     generational
    0.37
    POSITIVE LOGITS
     czyli
    0.88
     which
    0.80
     który
    0.75
     jossa
    0.73
     jonka
    0.70
    which
    0.70
     vilket
    0.70
     whiche
    0.69
     illetve
    0.68
     mutta
    0.68
    Activations Density 0.000%

    No Known Activations