© 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. 1995
    Prev
    Next
    INDEX
    Explanations

    Okay, let's break down

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

    This neuron strongly activates on sentence-initial or clause-linking words and punctuation (e.g. “Furthermore,” “Nevertheless,” “When,” “You,” commas), i.e. discourse-level transition markers that begin or link sentences.

    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
     불구
    0.54
    .:
    0.52
     再
    0.51
    }:$
    0.50
     ৫৭
    0.50
     הת
    0.50
    \":
    0.49
    WRITE
    0.49
     非
    0.49
    ㅈ
    0.49
    POSITIVE LOGITS
     However
    0.74
     you
    0.73
    あなたは
    0.71
    However
    0.70
     If
    0.67
     Inoltre
    0.66
     azonban
    0.66
     Allerdings
    0.63
     Međutim
    0.63
     Toutefois
    0.62
    Activations Density 0.831%

    No Known Activations