© 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. 31-GEMMASCOPE-2-RES-16K
    7. 6422
    Prev
    Next
    INDEX
    Explanations

    hesitation and filler words

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

    The neuron is highly activated by tokens that mark speaker turns or speaker labels (e.g. “S:”, “BROOKS:”, “CLINTON:”) in dialogue or transcript text.

    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
    ISO
    0.66
    ҡ
    0.55
    In
    0.50
    Arqu
    0.50
    يد
    0.50
    WX
    0.50
    Moreover
    0.49
    ISSN
    0.49
    WIT
    0.49
    ENA
    0.48
    POSITIVE LOGITS
     uh
    2.13
     um
    1.70
     ehm
    1.61
    呃
    1.53
    嗯
    1.48
     euh
    1.47
     Uh
    1.46
     เอ่อ
    1.38
     ähm
    1.34
     UH
    1.28
    Activations Density 0.197%

    No Known Activations