© 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. 6706
    Prev
    Next
    INDEX
    Explanations

    testing efficiency and performance

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

    The neuron strongly activates on the very first “content” token of each new text segment (i.e. it marks the start-of-segment token).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408

    words that immediately precede punctuation or special characters, particularly spaces before punctuation marks.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408

    clause-anchoring function words and quantifiers—along with key abstract nouns—marking the main focus of a sentence.

    oai_token-act-pair · gpt-5Triggered 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
    阙
    0.81
    weapp
    0.80
    tutorial
    0.79
    哪里
    0.79
     TextAppearance
    0.79
    mented
    0.77
    ウェア
    0.77
    ໄດ້
    0.73
    swear
    0.71
    ത്തക
    0.71
    POSITIVE LOGITS
     totale
    0.77
    ISE
    0.75
     арифмети
    0.75
     препара
    0.74
    ور
    0.73
     нау
    0.73
     Famil
    0.71
    ESE
    0.70
     દરમિયાન
    0.70
    во
    0.69
    Activations Density 0.000%

    No Known Activations