© 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 EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 40-GEMMASCOPE-2-TRANSCODER-262K
    4. 17660
    Prev
    Next
    INDEX
    Explanations

    This neuron is specifically sensitive to phrases indicating "young adults" or individuals in their late teens and early twenties, often associated with life transitions or specific age ranges (e.g., 18-25).Therefore, the most fitting explanation would relate to this demographic.`young adults` is a good candidate.`young and adult` is also good.`age range 18-25` is good too.Let's check the length requirement: 3-20 words.`young adults` (2 words) - fits.`young and adult` (3 words) - fits.`age range 18-25` (4 words) - fits.Observing the `MAX_ACTIVATING_TOKENS` again: `young`, `young`, `young`, `adult`.And `TOKENS_AFTER_MAX_ACTIVATING_TOKEN`: `woman`, `men`, `adulthood`, `son`.The primary and most consistent pattern points to the concept of "young adults".Final choice: `young adults` or `young person's age` or `young adult demographic`.young adults

    np_acts-logits-general · gemini-2.5-flash-lite
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-2-27b-it/transcoder_all/layer_40_width_262k_l0_small_affine
    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
    学生
    1.45
     college
    1.43
     студен
    1.41
     student
    1.37
    大学生
    1.37
    高校
    1.34
     دانشج
    1.34
    college
    1.31
     студент
    1.31
    學生
    1.30
    POSITIVE LOGITS
     miliardi
    0.42
    marshalN
    0.41
     Milliarden
    0.38
    avidin
    0.38
     thuyết
    0.38
     Gravel
    0.38
    pedo
    0.38
     jdbcTemplate
    0.38
     राजवंश
    0.37
    GTEST
    0.37
    Activations Density 0.086%

    No Known Activations