© 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 EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-12B
    3. 24-GEMMASCOPE-2-RES-16K
    4. 13278
    Prev
    Next
    INDEX
    Explanations

    patterns related to spreadsheet cell references and ranges, often found in input parameters or data definitions, particularly looking for cells or ranges indicated by 'A' followed by a number and/or dollar signs.Let's re-evaluate using the rules and the provided lists.- **MAX_ACTIVATING_TOKENS:** 'A'- **TOKENS_AFTER_MAX_ACTIVATING_TOKEN:** '$' - This means the neuron is activated when it sees 'A' followed by '$'.- **TOP_POSITIVE_LOGITS:** - ድረس: Arabic, "up to", "until", "limit" - inclusiv: "inclusive" (likely part of a range) - にかけて: Japanese, "up to", "until", "regarding" - 为止: Chinese, "until", "up to", "to the end" - तक: Hindi, "up to", "until" - límite: Spanish, "limit" - Limit: English, "limit" - 日まで: Japanese, "until day" - 일까지: Korean, "until day" - অবধি: Bengali, "until", "up to" - All these words point towards concepts of limits, ranges, or endpoints.- **TOP_ACTIVATING_TEXTS:** - "`A2`": This looks like a spreadsheet cell reference. - "`A$2:A$100`": This is a spreadsheet range reference. The dollar signs ($) are used for absolute referencing in spreadsheets.Combining these clues:- The activating token is 'A'.**spreadsheet cell references**

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

    The neuron predominantly activates on tokens that convey numeric or quantitative information (e.g. numbers, measurements, and related quantifiers).

    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
    case
    0.75
     inicia
    0.73
     earlier
    0.71
     started
    0.70
     पहली
    0.70
     başladı
    0.69
    Ảnh
    0.69
    र्जा
    0.68
     inizia
    0.68
     ડો
    0.66
    POSITIVE LOGITS
     ድረስ
    1.40
    日まで
    1.23
     وحتى
    1.18
    为止
    1.16
    まで
    1.16
     sampai
    1.10
     तक
    1.08
     aż
    1.08
    일까지
    1.06
     পর্যন্ত
    1.05
    Activations Density 0.769%

    No Known Activations