© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it
    Neuronpedia
    Jacobian LensNEW
    Natural Language
    Autoencoders
    MAY 26
    Assistant AxisJAN 26Circuit TracerAPR 26SteerSAE EvalsExportsinterp-engineNEW Community BlogPrivacy & TermsContact
    1. Home
    2. Gemma-3-27B-IT
    3. 24-GEMMASCOPE-2-TRANSCODER-262K
    4. 174767
    Prev
    Next
    INDEX
    Explanations

    Analyzing the data:- `MAX_ACTIVATING_TOKENS` predominantly features `sockaddr` and `addr`.- `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` shows `*)&` and `addr` following these tokens.- `TOP_ACTIVATING_TEXTS` contains numerous C/C++ code snippets related to networking, specifically using `struct sockaddr` and `(struct sockaddr *)&...` followed by `sizeof(...)` or variable assignments like `addr.sin_port`.- `TOP_POSITIVE_LOGITS` shows non-English words, which don't seem immediately related to the code context but can indicate the neuron's breadth or the underlying embedding space.The core pattern is clearly related to C/C++ networking structures

    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_24_width_262k_l0_small_affine
    Prompts (Dashboard)
    238,145 prompts, 512 tokens each
    Dataset (Dashboard)
    lmsys + oasst1
    No Configuration Found
    Embeds
    IFrame
    Link
    Negative Logits
     So
    0.46
    ;
    0.45
     or
    0.42
    a
    0.42
     so
    0.42
     ska
    0.40
    we
    0.40
    f
    0.39
    k
    0.39
     fü
    0.39
    POSITIVE LOGITS
    ליה
    0.45
     земля
    0.44
     প্রতিহিংস
    0.43
    ورسٹی
    0.43
     élections
    0.43
    驀
    0.42
    MODKEY
    0.41
     кабе
    0.41
     vipp
    0.41
    䢎
    0.41
    Activations Density 0.001%

    No Known Activations