© 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. 4727
    Prev
    Next
    INDEX
    Explanations

    used for emphasis

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

    The neuron fires on low-frequency or otherwise “unusual” tokens—things like proper names, technical jargon, and non-standard punctuation (e.g. em-dashes or quoted phrases).

    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
     
    0.66
    
    0.64
    
    0.64
     درمان
    0.63
     Kõ
    0.63
     caliper
    0.63
     rmse
    0.63
     Scho
    0.62
    attice
    0.62
    \}-
    0.62
    POSITIVE LOGITS
    这句话
    1.23
    这话
    1.13
     якобы
    0.93
     implying
    0.91
    presumably
    0.87
     কথাটা
    0.86
     apparently
    0.83
     offenbar
    0.82
     presumably
    0.81
     statement
    0.80
    Activations Density 0.534%

    No Known Activations