© 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-IT
    3. 12-GEMMASCOPE-2-RES-16K
    4. 14509
    Prev
    Next
    INDEX
    Explanations

    The neuron appears to be detecting dates, years, and numerical sequences within text. The examples show years like "1971," "2008," "1980," "1991," "1992," "1987," "1986," and decade indicators like "1980s." It also picks up numerical ranges like "25 to 40 million" and specific numbers like "28" and "3,000." The "MAX_ACTIVATING_TOKENS" also contains numerical data ("7", "8", "0", "1", "2", "3", "5", "6"), further supporting this observation.Therefore, a concise explanation capturing this behavior is:numbers and years

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-12b-it/resid_post/layer_12_width_16k_l0_medium
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    را
    0.92
    IN
    0.88
    P
    0.87
    IF
    0.80
    वा
    0.77
    রা
    0.76
    上海
    0.75
    Foi
    0.75
    B
    0.75
    AN
    0.75
    POSITIVE LOGITS
     ketchup
    0.75
     linebacker
    0.72
     plight
    0.72
     l
    0.71
     Because
    0.70
     abras
    0.67
     tireless
    0.67
     разными
    0.66
     CEOs
    0.66
     threats
    0.66
    Activations Density 0.059%

    No Known Activations