© 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
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 5807
    Prev
    Next
    INDEX
    Explanations

    **lists**- `TOP_POSITIVE_LOGITS`: `=`, `<`, `אן`, `inferences`, `shade`, `*$`, `しの`, `*`, `vitamins`, `רית`- `TOP_ACTIVATING_TEXTS`: - "off is September 2021, so I don't have information on events after that date." - "Summarization: I can summarize text," - "Drink a large glass of water right now. Keep a water bottle with you and sip throughout the day. * Move Around: Even a 5-10 minute walk can boost circulation and alertness. Stretch, do some" - "promoting free trade agreements (though this has become more nuanced recently with some Republicans favoring protectionist measures).I am unable to provide an explanation as the input is missing the `<MAX_ACTIVATING_TOKENS>` and `<TOKENS_AFTER_MAX_ACTIVATING_TOKEN>` sections, which are crucial for identifying specific token patterns. Without these, I cannot accurately determine the neuron's behavior

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

    Explanation of neuron 4 behavior: the main thing this neuron does is find numeric tokens—especially multi‐digit strings, timestamps, version numbers, or floating‐point values.

    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
     प्रसंस्करण
    0.53
     morphisms
    0.52
    様専用
    0.51
     Technologies
    0.49
    灝
    0.49
     करण्यासाठी
    0.48
     technologies
    0.48
     Enterprises
    0.48
     substrates
    0.48
     Sempre
    0.47
    POSITIVE LOGITS
     *$
    0.56
     `<`,
    0.52
     *
    0.51
     `:`,
    0.49
     Алек
    0.49
    tolower
    0.49
    FName
    0.49
    PersonalInfo
    0.48
    Fmat
    0.48
     тих
    0.47
    Activations Density 0.007%

    No Known Activations