© 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
    2. Gemma-3-12B
    3. 31-GEMMASCOPE-2-RES-16K
    4. 3064
    Prev
    Next
    INDEX
    Explanations

    This neuron fires on discourse markers and related tokens that introduce or highlight a contrast or concessive clause (e.g. “although,” “there are,” “few,” etc.).

    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
     warahmatullahi
    0.48
    GLIGENCE
    0.47
     severally
    0.46
    imony
    0.46
    ట్లో
    0.46
     단순히
    0.45
    isPartOf
    0.44
    芘
    0.44
    гии
    0.44
    Facades
    0.43
    POSITIVE LOGITS
     new
    5.69
    new
    5.13
    新的
    5.03
     nuevos
    4.97
     neuen
    4.91
     новых
    4.91
     নতুন
    4.88
     nieuwe
    4.88
     nuovi
    4.88
     новый
    4.84
    Activations Density 1.867%

    No Known Activations