© 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. 10-GEMMASCOPE-2-TRANSCODER-262K
    4. 205752
    Prev
    Next
    INDEX
    Explanations

    buzzfeed quizes, lists and articlesThis neuron is strongly activated by content related to BuzzFeed, specifically its quizzes, lists, and articles. The `MAX_ACTIVATING_TOKENS` show "BuzzFeed" as a high-frequency token. The `TOKENS_AFTER_MAX_ACTIVATING_TOKEN` list shows tokens like "Code" and "Reduction" appearing after these activating tokens in some texts, which might imply how BuzzFeed's content relates to generating code or simplifying information, or it could be coincidental noise from the longer text examples. However, the presence of "BuzzFeed" itself, along with numerous examples of "Feed" and "plate" (which can be related to feeding content or information), strongly points towards content platforms like BuzzFeed. The `TOP_POSITIVE_LOGITS` are mostly non-English characters and symbols, which suggests the neuron might be separating languages or types of characters, but the core pattern revolves around the content type indicated by "BuzzFeed".Therefore, a concise explanation focusing on this primary pattern would be:**BuzzFeed quizzes, lists and articles**

    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_10_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
    ied
    0.81
     for
    0.75
    du
    0.72
    dog
    0.70
    des
    0.66
    dT
    0.66
    dil
    0.65
    ung
    0.64
    ok
    0.64
    di
    0.64
    POSITIVE LOGITS
    ى
    0.85
    幵
    0.78
    ون
    0.71
     μεγά
    0.70
     男女
    0.70
    <0x80>
    0.67
    县
    0.67
     .\
    0.66
     unters
    0.63
     werde
    0.63
    Activations Density 0.000%

    No Known Activations