© 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. 2329
    Prev
    Next
    INDEX
    Explanations

    care, focus, impl, step, special

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

    The neuron fires strongly on words that begin sentences or speaker turns—i.e. capitalized tokens at the start of a new utterance or line.

    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
     (
    0.66
                    
    0.61
                 
    0.58
    <eos>
    0.57
                       
    0.57
    ↵
    0.56
                                   
    0.56
    <0xE2>
    0.56
                
    0.55
     "
    0.55
    POSITIVE LOGITS
    прочем
    0.82
     übrigens
    0.76
    laublich
    0.71
    heureusement
    0.70
    ielleicht
    0.69
     правда
    0.68
     kanske
    0.66
     těch
    0.66
    何况
    0.66
    ratulations
    0.65
    Activations Density 0.810%

    No Known Activations