© 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-3-12B
    5. 31-GEMMASCOPE-2-RES-16K
    7. 11726
    Prev
    Next
    INDEX
    Explanations

    bacardi, baked, baking

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

    The neuron strongly activates on three‐letter subword prefixes (e.g. “vac-” in vacuum/vacation, “bak-” in bake/Baker, “baq-” in Baqara), i.e. three‐character stems at the start of words.

    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
    ﹏
    0.55
    𝔞
    0.54
     Grower
    0.52
    𝕠
    0.52
     জানুয়ার
    0.51
     запас
    0.51
     Empres
    0.50
    ु
    0.49
    quick
    0.49
    Circular
    0.49
    POSITIVE LOGITS
     Vak
    0.69
     vacuum
    0.68
     Bak
    0.65
     vacuoles
    0.65
     vac
    0.64
     bac
    0.64
     vacation
    0.64
     vak
    0.64
    vak
    0.64
     BAK
    0.64
    Activations Density 0.219%

    No Known Activations