© 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
    2. Gemma-3-12B
    3. 31-GEMMASCOPE-2-RES-16K
    4. 8348
    Prev
    Next
    INDEX
    Explanations

    attributing speech

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

    The neuron fires on reported‐speech cues—i.e. quotation marks and speech‐attribution verbs like “said,” “says,” and “told.”

    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
    💬
    0.77
     কিছুটা
    0.69
     مزید
    0.68
     Cllr
    0.68
    further
    0.68
     excitedly
    0.68
    सुद्धा
    0.67
     sottoline
    0.66
    បង្ហ
    0.65
    Cllr
    0.65
    POSITIVE LOGITS
     =
    0.61
     ==
    0.61
     =-
    0.55
     Happened
    0.55
     !=
    0.54
    '-
    0.54
    <0xE3>
    0.53
     th
    0.53
    <unused2174>
    0.53
     ​​​​
    0.53
    Activations Density 0.041%

    No Known Activations