© 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. 24-GEMMASCOPE-2-RES-16K
    7. 7558
    Prev
    Next
    INDEX
    Explanations

    cannot and will not

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

    The neuron activates on first‐person evaluative or emphatic expressions (e.g. “can’t,” “cannot,” “couldn’t,” “believe,” “recommend,” “begin to”) that signal a strong personal reaction or endorsement.

    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
    '
    1.08
    (
    1.06
    "
    0.94
    0.86
    ও
    0.85
     lahko
    0.82
    want
    0.80
    0
    0.80
    wil
    0.80
    hibition
    0.79
    POSITIVE LOGITS
     videomuzda
    0.93
     ஆகவே
    0.93
    𐰆
    0.90
    ट्टर
    0.87
    з
    0.85
    ⿷
    0.84
    ᖕ
    0.83
    $-$,
    0.82
    től
    0.82
    ज़न
    0.82
    Activations Density 0.088%

    No Known Activations